fix: close what nine adversarial reviews found, most of it mine

Nine focused reviews (export state machine, upload path, auth/abuse, client
queue, guest UI, database, deploy/ops, regression hunt, test honesty). Every
finding below was re-verified against the code before being acted on; several
plausible-sounding ones were checked and rejected.

## Data loss and denial of service

**One request could OOM-kill the app container.** `client_upload_id` was read
with `Field::text()` — axum builds its multipart reader with no SizeLimit, so
the only bound was the route's 576 MiB body limit, then decoded into a second
full String. `caption` and `hashtags` go through `read_text_field_bounded` for
exactly this reason; this field arrived later and missed it. Any guest, one
request, and every SSE stream drops and every in-flight temp file is stranded.

**Nothing bounded concurrent upload bodies.** The headroom gate can only refuse
to COMMIT — the body is already streamed to a temp file by the time it runs, and
neither axum, the tower stack nor Caddy limits how many stream at once. ~100
guests tapping "upload all" after the ceremony puts 10-20 GB of .tmp on a 40 GB
volume, invisible to the gate, eating the reserve that keeps Postgres able to
write WAL. New `UploadAdmission` budgets bytes (not requests, so one video and
two hundred photos coexist) via a permit that releases on drop, so every exit
path returns it.

**The export decode bypassed the memory permit the compression path takes.**
Same class of work — decode + resize every image in the gallery — in a bare
spawn_blocking. A release fired while the last photos were still compressing put
both in the same 1 GiB cgroup; the OOM kill marks the export failed and
`recover_exports` re-spawns it into the same conditions on the next boot. The
permit is now process-wide in `imaging`, because the constraint it expresses is
the container's memory, not one worker's.

**`MediaTotalCache` cached its own failure as 0.** For the whole TTL the gate
then saw an empty event and collapsed to the flat reserve — the behaviour the
two-halves design replaced — with no log line. And the trigger correlates with
the danger: with max_connections 10 the query fails exactly during a burst. Now
falls back to the last good reading and says so.

**V8's heap ceiling sat above the frontend container's entire budget** (measured:
259 MB inside a 256M limit), so GC could never intervene and the only
backpressure was SIGKILL under an arrival burst.

## Guest-visible

**The feed stopped being newest-first after the first reconcile.** It fetches
whole 100-item server pages while `uploads` grows in 20s, so everything in the
gap was absent from `present`, classified as new, and prepended — ~80 photos
from earlier in the evening above the newest ones. It also stalled infinite
scroll, since the cursor still pointed at item 20 and the observer only re-fires
on a change. The union is now sorted on the server's own (created_at, id) key,
which additionally places an SSE arrival correctly.

**A stale `loadMoreError` outlived every refresh and filter change**, leaving a
false error above a button that returns immediately on `!nextCursor`.

**A failed derivative toasted "Ein Upload konnte nicht verarbeitet werden."** for
a photo sitting right there on screen — the handler still assumed 1d9fb11's
pre-fix behaviour (row deleted, quota refunded, card evicted), none of which is
true any more. It was the last surviving route for the "your photo is gone"
signal that fix set out to remove.

## Enforcement that existed only in comments

`recover_name_rate_per_15min` is clamped at the point of use: the ordering
`3 x ceiling <= PIN_LOCK_THRESHOLD` is the whole control against one source
locking any guest whose name is on the feed, it was asserted in a comment, and
`patch_config` accepted 1..100_000. The test pinned the default constant rather
than the enforced bound; it now pins the bound.

## Tests that could not fail

- The gate test asserted only its own premise (`500MB x 100 > 35GB`) and never
  touched the gate. It now checks both controls against the same state and
  requires them to disagree in the right direction.
- `the_banner_always_fires_before_the_upload_gate_closes` reduced to
  `G < G + G/4` — true for any margin, including zero, so it could not detect
  the banner moving to exactly the gate. It now pins the gap.
- `disk_is_low`'s `free < LOW_DISK_FLOOR_BYTES` clause was unreachable (warn_at
  is always >= 12.5 GB against a 10 GB floor). Two tests were named after it and
  neither could fail if it were deleted. Clause and constant removed.
- The suspension test I added last commit hard-coded the credit cap instead of
  importing it, so changing STALL_TIMEOUT_MS would leave it passing against a
  system that no longer exists. Now imports MAX_SUSPEND_CREDIT_MS.

## Stale comments corrected

The prune doc still argued at length for the pre-build ordering that 1d9fb11
reversed — a reader trusting it would reopen the blocker 0506369 fixed.
DISK_RESERVE_BYTES claimed to equal the banner threshold that 0506369
deliberately offset by 25%. And host.rs kept its own duplicate 10 GB literal
instead of importing the constant.

154/154 backend, 59/59 vitest, clippy clean, svelte-check 0 errors, eslint
clean, both builds, compose + caddy validate.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

.claude/settings.json

@@ -0,0 +1,9 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(cargo check *)",
+      "Bash(cargo clippy *)",
+      "Bash(git --no-pager diff *)"
+    ]
+  }
+}

.env.example

@@ -1,7 +1,19 @@
 # ── Domain ────────────────────────────────────────────────────────────────────
 # Public domain Caddy will serve and obtain a TLS certificate for.
+#
+# The DNS A record must already point at this server BEFORE the first `up -d`: Caddy
+# requests a certificate on boot, and Let's Encrypt allows only 5 failed validations per
+# hostname per hour. Never delete the caddy_data volume — it holds the certificate and
+# the ACME account key.
 DOMAIN=my-event.example.com
 
+# ── Image version ─────────────────────────────────────────────────────────────
+# Tag pulled for the `app` and `frontend` services (docker-compose.yml). Production runs
+# prebuilt images from the registry and never compiles — see DEPLOYMENT_RUNBOOK.md.
+# Always an immutable tag, never `latest`: rollback is `EVENTSNAP_VERSION=<previous>`
+# + `docker compose up -d`, which works offline if that image is still resident locally.
+EVENTSNAP_VERSION=v0.13.0
+
 # ── App server ────────────────────────────────────────────────────────────────
 APP_PORT=3000
 # Set to `production` in real deployments. This activates the secret guard that
@@ -12,10 +24,39 @@ APP_ENV=production
 # ── Database ──────────────────────────────────────────────────────────────────
 # Set a strong password and keep it in sync between DATABASE_URL and
 # POSTGRES_PASSWORD. Generate one with: openssl rand -hex 24
+#
+# SET THIS BEFORE THE FIRST `docker compose up -d`. Postgres reads POSTGRES_PASSWORD
+# only when it initialises its data directory, on that very first boot. Change it
+# afterwards and the app authenticates with the new password against a volume still
+# holding the old one — a permanent restart loop ("password authentication failed").
+# The only ways out are restoring the old password or `docker compose down -v`, which
+# deletes the database, the media and the exports. In production the app refuses to
+# boot while this is still the placeholder below, so it cannot be missed by accident.
 DATABASE_URL=postgres://eventsnap:CHANGE_ME_use_a_strong_password@db:5432/eventsnap
 POSTGRES_USER=eventsnap
 POSTGRES_PASSWORD=CHANGE_ME_use_a_strong_password
 POSTGRES_DB=eventsnap
+# Connection pool size. The code default is 10 (backend/src/db.rs) — set it explicitly,
+# because a `.env` written by hand from this file's secrets is otherwise silently on 10.
+#
+# SIZE IT TO THE CORES, NOT TO THE GUESTS. The earlier advice here was ~30, reasoned from
+# "~100 guests polling the feed at once" back when a feed page cost ~449 ms and connections
+# were spent waiting. Migration 024 replaced the feed view's GROUP BY with scalar subqueries
+# and a page now costs well under a millisecond, so concurrency is no longer where the time
+# goes. On a 2 vCPU box 30 simultaneous queries cannot run — they queue on the CPU instead of
+# on the pool, which is the same wait wearing a different hat, and 30 Postgres backends plus
+# shared_buffers is snug in the 1G that docker-compose.yml allots `db`.
+#
+# 15 on 2 vCPU / 4 GB. Raise toward 30 only alongside more cores AND a bigger `db` memory
+# limit — an OOM in Postgres doesn't degrade one feature, it takes the whole event down.
+DATABASE_MAX_CONNECTIONS=15
+
+# Log level: see the "Logging" section near the bottom of this file.
+#
+# Defined THERE and nowhere else, deliberately. This file used to assign RUST_LOG twice —
+# once here and once there — and Compose takes the LAST assignment, so editing this line to
+# `debug` to chase a problem during the event changed nothing at all, silently. A key that
+# appears twice in a .env is a trap regardless of which value is better.
 
 # ── Authentication ────────────────────────────────────────────────────────────
 # Generate with: openssl rand -hex 64
@@ -23,8 +64,14 @@ JWT_SECRET=change_me_to_a_random_64_byte_hex_string
 SESSION_EXPIRY_DAYS=30
 
 # Admin dashboard password (bcrypt hash).
-# Generate with: htpasswd -bnBC 12 "" yourpassword | tr -d ':\n'
-ADMIN_PASSWORD_HASH=$2y$12$placeholder_replace_me
+# Generate with an image the stack already pulls (htpasswd needs apache2-utils, which
+# a stock VPS does not have):
+#   docker run --rm caddy:2-alpine caddy hash-password --plaintext 'yourpassword'
+# IMPORTANT: keep the SINGLE QUOTES. A bcrypt hash is full of `$` (e.g. $2b$12$…$…),
+# and both Docker Compose's env_file interpolation and dotenvy's variable substitution
+# would otherwise eat the `$…` segments (reading them as unset vars) and corrupt the
+# hash — every admin login then 401s. Single quotes make both read it literally.
+ADMIN_PASSWORD_HASH='$2y$12$placeholder_replace_me'
 
 # ── Event ─────────────────────────────────────────────────────────────────────
 EVENT_NAME=Max & Maria's Wedding
@@ -36,19 +83,97 @@ MEDIA_PATH=/media
 # /media is publicly served, so exports here would be downloadable without auth.
 EXPORT_PATH=/exports
 
-# ── Upload limits ─────────────────────────────────────────────────────────────
-DEFAULT_MAX_IMAGE_SIZE_MB=20
-DEFAULT_MAX_VIDEO_SIZE_MB=500
-
-# ── Rate limiting ─────────────────────────────────────────────────────────────
-DEFAULT_UPLOAD_RATE_PER_HOUR=10
-DEFAULT_FEED_RATE_PER_MIN=60
-DEFAULT_EXPORT_RATE_PER_DAY=3
-
-# ── Capacity ──────────────────────────────────────────────────────────────────
-DEFAULT_ESTIMATED_GUEST_COUNT=100
-# Fraction of total storage that triggers the "low storage" warning (0.0–1.0)
-DEFAULT_QUOTA_TOLERANCE=0.75
+# ── Runtime settings (upload limits, rate limits, capacity) ───────────────────
+# NOTE: These are NOT environment variables. Upload size caps, rate limits, guest
+# count and quota tolerance are stored in the database `config` table (seeded once
+# at first boot) and changed at runtime from the ADMIN DASHBOARD — the backend does
+# not read them from .env. Setting them here has no effect. Current seeded defaults:
+#   upload rate      100 / hour / guest   (raised from 10 by migration 015)
+#   feed rate         60 / minute
+#   export rate        3 / day
+#   max image size    20 MB
+#   max video size   500 MB
+#   estimated guests 100
+#   quota tolerance  0.75  (see below — NOT a warning threshold)
+# Adjust these in the admin UI before the event if needed.
+#
+# quota_tolerance is the MULTIPLIER IN THE PER-USER QUOTA FORMULA, not the point at
+# which anything warns you:
+#
+#     per_user_limit = floor(free_disk * quota_tolerance / active_uploaders)
+#
+# It is recomputed against LIVE free space on every upload, so it self-throttles: guests
+# converge on a fixed point at tolerance/(1+tolerance) of the free space you started
+# with — 43% at 0.75, i.e. ~30 GB of a fresh 70 GB.
+#
+# Raising it therefore AUTHORISES GUESTS TO FILL MORE OF THE DISK. Setting 0.95 in the
+# belief that it means "warn me later" moves the fixed point to ~49% and eats the
+# headroom the keepsake needs — and the keepsake needs a lot, because Gallery.zip and
+# Memories.zip are each roughly a second copy of every original (both store media
+# uncompressed). Budget for media + 2x media, or move exports to their own volume.
+#
+# 0.75 is the tested default. Lower it if the box is tight; raise it only if you have
+# provisioned export headroom separately.
 
 # ── Workers ───────────────────────────────────────────────────────────────────
+# Number of parallel media compression workers. Default 2. Boot-time only.
+#
+# CORRECTION TO EARLIER GUIDANCE: this used to say "each worker can run an ffmpeg
+# transcode, so raise the app memory limit to ~2G if you set 4". There is NO video
+# transcode anywhere in this codebase — services/video.rs runs
+# `ffmpeg -ss <t> -i <src> -vframes 1 -vf scale=...`, a single poster frame, and video
+# originals are stored and served byte-for-byte. Poster extraction costs ~150-250 MB
+# for a moment; it is not the constraint.
+#
+# The real memory consumer is the IMAGE path. `image` 0.25's resize builds an Rgba32F
+# intermediate at 16 BYTES PER PIXEL, sized (source_width x target_height) — which the
+# 256 MiB decode guard in imaging.rs does NOT cover. Peak per photo, decode + the 2048px
+# display resize: ~145 MB at 12 MP, ~223 MB at 24 MP, ~354 MB at 48 MP.
+#
+# So on a 2 vCPU / 4 GB box (e.g. Hetzner CX22) KEEP THIS AT 2:
+#   * concurrency 4 would put two giants at ~1.5 GB against the 1G app limit — OOM.
+#   * and app=2G + db=1G + frontend/caddy 256M each + ~370 MB of OS/Docker exceeds the
+#     ~3910 MiB a "4 GB" VM actually reports. Raising the limit oversubscribes the host.
+# 4 is only reasonable on the 4 vCPU / 8 GB box README.md documents.
+#
+# The "two 48 MP photos at once" worst case this number used to be sized against is no
+# longer reachable: compression.rs takes an EXCLUSIVE `heavy` permit for any job whose
+# estimated peak exceeds HEAVY_IMAGE_BYTES (150 MiB), so two giants serialise no matter what
+# this is set to. What concurrency 2 now buys is two ORDINARY phone photos in parallel
+# (~145 MB peak each), which is both memory-safe and short enough not to starve the two
+# tokio worker threads a 2 vCPU box gets.
+#
+# Do NOT drop this to 1 hoping to protect the CPU. It halves throughput on the common light
+# path for a heavy path that is already serialised, and a longer compression backlog means
+# more feed tiles served from full-size originals (VirtualFeed falls back to /original while
+# derivatives are pending) — trading a little CPU for a lot of venue-wifi bandwidth.
+#
+# Throughput at 2 is not the bottleneck anyone thinks it is: ~2.5s per 12 MP photo, so
+# 100 photos is ~250 CPU-seconds spread over an entire evening.
 COMPRESSION_WORKER_CONCURRENCY=2
+
+# ── Comments ──────────────────────────────────────────────────────────────────
+# Master switch for the comment feature. Boot-time only (NOT in the admin UI), so it
+# needs a `docker compose up -d` to apply. Anything other than false/0/no/off is on.
+#
+# When false the backend rejects NEW comments with 403 and the frontend hides the whole
+# comment UI, including in the offline keepsake viewer. Likes and captions are entirely
+# separate features and are unaffected. Existing comments stay in the database (hidden),
+# so flipping it back restores them.
+#
+# Note it gates POSTING only: GET /upload/{id}/comments still serves already-existing
+# comments, and the keepsake's data.json still embeds their text. Irrelevant if the flag
+# is off from the first boot, since no comment can ever have been written.
+# NOTE for the current deployment: `docker-compose.yml` PINS this to "false" on the app
+# service, and `environment` overrides `env_file` — so changing it here has no effect in
+# production. Remove that line from the compose file first if you want comments back.
+COMMENTS_ENABLED=true
+
+# ── Logging ───────────────────────────────────────────────────────────────────
+# SET THIS IN PRODUCTION. Without it the app falls back to
+# `eventsnap_backend=debug,tower_http=debug` (see main.rs), and with TraceLayer that is a
+# debug line per HTTP request — including every preview and thumbnail fetch. Combined with
+# Docker's json-file driver it writes to the same filesystem as the database and the media.
+# docker-compose.yml caps each service's logs at 30 MB; this keeps the volume sane in the
+# first place. The e2e stack has always used exactly this value.
+RUST_LOG=eventsnap_backend=info,tower_http=warn

.github/workflows/e2e.yml

@@ -7,7 +7,7 @@ on:
 
 jobs:
   e2e:
-    name: Playwright E2E (chromium-desktop)
+    name: Playwright E2E (chromium + webkit)
     runs-on: ubuntu-latest
     timeout-minutes: 30
     steps:
@@ -25,7 +25,7 @@ jobs:
 
       - name: Install Playwright browsers
         working-directory: ./e2e
-        run: npx playwright install --with-deps chromium
+        run: npx playwright install --with-deps chromium webkit
 
       - name: Bring up the test stack
         working-directory: ./e2e
@@ -54,6 +54,22 @@ jobs:
         working-directory: ./e2e
         run: npm run test:e2e -- --project=chromium-mobile
 
+      # iOS Safari is the app's stated primary user (a wedding guest opening a QR link), and
+      # WebKit is the ONLY engine here that reproduces two of its behaviours:
+      #   - it enforces X-Frame-Options on the download iframe, so a site-wide `DENY` makes the
+      #     keepsake download silently do nothing. Blink hands attachments to the download
+      #     manager first and never notices. That shipped once already.
+      #   - it abandons a <video> load without a 206 response to its Range probe.
+      # Both regressions are invisible to every Chromium project, so running WebKit is what
+      # actually gates them on a PR rather than on someone remembering to test locally.
+      #
+      # The project is scoped in playwright.config.ts to the journeys a guest walks
+      # (01-auth, 02-upload, 03-feed, 06-export); four IndexedDB-blob tests skip themselves
+      # there — see helpers/webkit.ts for why that is the harness and not the app.
+      - name: Run E2E tests (webkit / iOS)
+        working-directory: ./e2e
+        run: npm run test:e2e -- --project=webkit-iphone
+
       - name: Upload Playwright report
         if: failure()
         uses: actions/upload-artifact@v4

.gitignore

@@ -13,8 +13,16 @@ frontend/build/
 frontend/export-viewer/node_modules/
 frontend/export-viewer/.svelte-kit/
 
-# Media uploads (mounted volume in production)
-media/
+# Media uploads. In production these live in the `media_data` DOCKER VOLUME, never in the
+# working tree — so this pattern is anchored to the repo root and exists only for a local
+# bind-mount experiment.
+#
+# It used to read `media/`, unanchored, which matches a directory of that name at ANY depth.
+# The only one in the repo is `e2e/fixtures/media/`, so the rule's entire practical effect was
+# to keep every E2E fixture untracked: a fresh clone got the specs and none of the images or
+# videos they read. `.github/workflows/e2e.yml` does a plain checkout and generates nothing, so
+# the committed CI job could not have run the upload, video or export suites at all.
+/media/
 
 # Playwright E2E suite — runtime artifacts (the suite itself is committed)
 e2e/node_modules/
@@ -29,3 +37,6 @@ e2e/.env.test
 # OS
 .DS_Store
 Thumbs.db
+
+# Claude Code personal (per-user) settings — shared settings.json IS committed
+.claude/settings.local.json

Caddyfile

@@ -9,34 +9,125 @@
 	header {
 		Strict-Transport-Security "max-age=31536000; includeSubDomains"
 		X-Content-Type-Options "nosniff"
-		X-Frame-Options "DENY"
 		Referrer-Policy "strict-origin-when-cross-origin"
 	}
 
+	# X-Frame-Options: DENY everywhere EXCEPT the keepsake download endpoints, which
+	# are navigated in a HIDDEN, SAME-ORIGIN iframe so a 404/429 can't unload the PWA
+	# (see frontend/src/routes/export/+page.svelte). WebKit enforces XFO *before*
+	# honouring Content-Disposition, so a blanket DENY makes the download silently do
+	# nothing on iOS Safari — the app's primary platform. SAMEORIGIN still blocks
+	# cross-origin framing.
+	#
+	# Split into two disjoint matchers rather than an override: Caddy applies the
+	# FIRST header directive outermost, so it wins on write — a later, more specific
+	# `header` would be silently ignored.
+	@framable path /api/v1/export/zip /api/v1/export/html
+	@not_framable not path /api/v1/export/zip /api/v1/export/html
+	header @framable X-Frame-Options "SAMEORIGIN"
+	header @not_framable X-Frame-Options "DENY"
+
 	# SvelteKit frontend — static assets with long-lived cache (content-hashed filenames)
 	@hashed_assets path_regexp hashed /_app/immutable/.*\.[a-f0-9]{8,}\.(js|css|woff2)$
 	header @hashed_assets Cache-Control "public, max-age=31536000, immutable"
 
-	# Preview/thumbnail images. These are now served by the app through a
-	# visibility-checked alias (/api/v1/upload/{id}/{preview,thumbnail}) so moderation
-	# can revoke access; direct /media/previews|thumbnails is 404-blocked at the app.
-	# Privately cacheable for a short window (the app sets the same header; this is the
-	# edge carve-out from the blanket no-store below). Kept short so a moderated image
-	# stops being served to a direct-URL holder promptly.
-	@media_api path /api/v1/upload/*/preview /api/v1/upload/*/thumbnail
+	# Preview/thumbnail/display images. These are served by the app through a
+	# visibility-checked alias (/api/v1/upload/{id}/{preview,thumbnail,display}) so
+	# moderation can revoke access; the app serves no /media route at all, so there is no
+	# direct path to the bytes. Privately cacheable for a short window (the app sets the
+	# same header; this is the edge carve-out from the blanket no-store below). Kept short
+	# so a moderated image stops being served to a direct-URL holder promptly.
+	#
+	# `display` was missing here while the backend set `private, max-age=300` on it, and
+	# because `header` REPLACES, the blanket no-store below silently won. That route is the
+	# ~2048px derivative the diashow uses exclusively, so a projector left running all
+	# evening re-fetched a full-size JPEG for every slide — roughly 2-4 GB pulled through
+	# the app over 8 hours, on the same venue uplink 100 guests are uploading over, and a
+	# blank frame on every network hiccup.
+	@media_api path /api/v1/upload/*/preview /api/v1/upload/*/thumbnail /api/v1/upload/*/display
 	header @media_api Cache-Control "private, max-age=300"
 
-	# API — never cache, EXCEPT the gated image routes above.
+	# API and health — never cache, EXCEPT the gated image routes above. A cached health
+	# response would report the last known state rather than the current one.
 	@api {
-		path /api/*
-		not path /api/v1/upload/*/preview /api/v1/upload/*/thumbnail
+		path /api/* /health
+		not path /api/v1/upload/*/preview /api/v1/upload/*/thumbnail /api/v1/upload/*/display
 	}
 	header @api Cache-Control "no-store"
 
-	# Route API and media requests to the Rust backend
+	# Route API and media requests to the Rust backend.
+	#
+	# The app serves no /media route at all (see the note in backend/src/main.rs) — media
+	# bytes are reachable only through the visibility-checked /api/v1/upload aliases, so
+	# /media/* forwards to a plain 404. The proxy line is kept deliberately: it means the
+	# edge faithfully hands /media to the app, so if a future change ever re-introduces a
+	# static media route the e2e gating specs see it here exactly as production would,
+	# instead of being masked by the SvelteKit 404 page.
 	reverse_proxy /api/* app:3000
 	reverse_proxy /media/* app:3000
 
+	# The backend registers /health on its ROOT router, not under /api/v1, so it needs its
+	# own line — without it the catch-all below hands /health to SvelteKit, which has no
+	# such route and returns its 404 page. That made the documented post-deploy check
+	# (`curl -fsS https://DOMAIN/health`) fail 100% of the time on a perfectly healthy
+	# stack. e2e/Caddyfile.test has always carried this line; production never did.
+	reverse_proxy /health app:3000
+
 	# Everything else goes to SvelteKit frontend
 	reverse_proxy frontend:3001
+
+	# Last-resort page for when Caddy itself cannot reach an upstream — the app or frontend
+	# container down, restarting, or still warming up after a host reboot. Without it a guest
+	# gets Caddy's bodiless 502: a completely blank page, which reads as "the whole thing is
+	# gone" rather than "try again in a moment".
+	#
+	# THIS DOES NOT TOUCH APPLICATION ERRORS. `handle_errors` fires only on errors CADDY
+	# generates; a status the app returns through `reverse_proxy` is written back verbatim and
+	# never reaches here. That distinction is load-bearing rather than incidental: the keepsake
+	# download navigates a HIDDEN IFRAME and depends on a real 404/429 arriving from the app
+	# (frontend/src/routes/export/+page.svelte), and every API route answers 403/404/429 as
+	# ordinary JSON that the client parses. Swallowing those into an HTML page would be a far
+	# worse regression than the blank 502 this fixes. Verified against this exact config: an
+	# upstream 404 through `reverse_proxy` still arrives as `Content-Type: application/json`
+	# with its body intact, while only a dial failure renders the page below.
+	#
+	# Scoped to 5xx so a hypothetical future Caddy-generated 4xx (there is none today) still
+	# returns plainly instead of claiming the server is restarting.
+	#
+	# The body is inline because the caddy service mounts ONLY ./Caddyfile and caddy_data —
+	# there is no volume to ship an HTML file through and the image has no build step, so a
+	# static file would mean changing the deployed stack's compose definition. No external
+	# font, stylesheet or image is referenced: the app may be exactly what is down.
+	#
+	# `handle_errors` has NO position in the directive order — Caddy hoists it into a separate
+	# `errors` route list — so it cannot disturb the "first `header` directive wins" hazard
+	# documented at the top of this file. The site-wide security headers still apply to it.
+	handle_errors 5xx {
+		header Content-Type "text/html; charset=utf-8"
+		header Cache-Control "no-store"
+		# {err.status_code} preserves the real status. Hardcoding 503 would mislabel a genuine
+		# 502 for anything watching from outside.
+		respond `<!doctype html>
+<html lang="de">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>Gleich zurück</title>
+<style>
+html{background:#faf9f7;color:#1a1918;font-family:system-ui,-apple-system,"Segoe UI",Roboto,sans-serif}
+body{margin:0;min-height:100vh;display:flex;align-items:center;justify-content:center;padding:2rem;text-align:center}
+h1{font-family:Georgia,"Times New Roman",serif;font-weight:600;font-size:1.5rem;margin:0 0 .75rem}
+p{margin:0;color:#545350;line-height:1.5}
+@media (prefers-color-scheme:dark){html{background:#100f0f;color:#f5f4f2}p{color:#a6a4a1}}
+</style>
+</head>
+<body>
+<main>
+<h1>Wir sind gleich zurück</h1>
+<p>Die Seite wird gerade neu gestartet.<br>Bitte lade in einem Moment neu — deine Fotos bleiben gespeichert.</p>
+</main>
+</body>
+</html>
+` {err.status_code}
+	}
 }

DEPLOYMENT_RUNBOOK.md

@@ -0,0 +1,693 @@
+# EventSnap — Production Deployment Runbook
+
+Target: **Hetzner CX22** (2 vCPU, 4 GB RAM, 40 GB disk), single host, docker compose.
+Event profile: ~100 guests, ~100 photos + a few videos, one evening, **operator attending and unavailable to troubleshoot**.
+
+Everything here is written for that last constraint. Where a choice trades throughput for
+"cannot need attention on the night", it takes the stable option.
+
+> **Note on this repo's own docs.** `README.md:62` and `PROJECT.md:359` specify a **CX33
+> (4 vCPU / 8 GB / 80 GB)**. You are deploying to half of that on every axis. Two pieces of
+> tuning advice in `.env.example` are calibrated for the CX33 and are actively wrong for a
+> CX22 — they are called out in §3. Trust this file over `.env.example` for sizing.
+
+---
+
+## 0. Timeline — the single most important control
+
+### Step zero: commit everything, before you build anything
+
+**The production deployment does not exist in git yet.** At the time of writing, `docker-compose.yml`
+and `.env.example` are modified but uncommitted, and `DEPLOYMENT_RUNBOOK.md`,
+`docker-compose.build.yml`, both `.dockerignore` files and migrations `021`/`022` are untracked.
+None of them is gitignored — they are simply not committed.
+
+That is not a tidiness problem, it is the deployment failing in two ways at once:
+
+- §7 tells you to `git clone` onto the server. `git show HEAD:docker-compose.yml` still has
+  `build:` keys and **no `image:` keys**, so on that clone `docker compose pull` skips both services
+  and `docker compose up -d` starts **a fat-LTO release build of 427 crates on the CX22** — the
+  exact scenario §1 rules out as an expected OOM. The committed file also has no log rotation.
+- `sqlx::migrate!()` embeds `./migrations` **at compile time**. A build from the working tree bakes
+  in 021 and 022 and applies them on first boot; any later rebuild from a clean clone produces an
+  image that lacks them and crash-loops with `VersionMissing` against its own database.
+
+```bash
+git add docker-compose.yml docker-compose.dev.yml docker-compose.build.yml .env.example \
+        backend/.dockerignore frontend/.dockerignore frontend/Dockerfile \
+        backend/migrations/021_*.sql backend/migrations/022_*.sql \
+        DEPLOYMENT_RUNBOOK.md
+git commit -m "chore: production compose, ignore files and runbook"
+git push
+
+# Prove it landed — this must print two `image:` lines and nothing about `build:`
+git show HEAD:docker-compose.yml | grep -E 'image:|build:'
+git show HEAD --stat | grep -c migrations/02   # must be 4
+```
+
+| When | What |
+|---|---|
+| **T‑7 days** | Commit and push everything above. Registry + DNS pre-flight (§5). Build and push images (§6). |
+| **T‑5 days** | First deploy to the server (§7). Verify admin login. Leave it running. |
+| **T‑3 days** | **Freeze migrations.** No further code deploys unless something is broken. |
+| **T‑2 days** | Pre-pull current *and* previous image tags (§9). Run the backup rehearsal (§10). |
+| **Event day** | Change nothing. Configuration tweaks via the admin dashboard only (§4). |
+
+**Why the freeze matters more than anything else here.** Migrations run automatically at boot
+(`backend/src/db.rs`, `create_pool`) and `sqlx::migrate!()` is used *without* `set_ignore_missing`. An older
+image booted against a newer schema does not degrade — it **crash-loops** with `VersionMissing`.
+The repo documents this itself in `backend/migrations/014_export_epoch.up.sql:3-8`. Once the
+migration set is frozen, rollback is a one-line `.env` edit; before it is frozen, rollback means
+hand-running down-SQL under pressure.
+
+---
+
+## 1. Deployment strategy — build on the Mac, push, pull
+
+**Decision: build `linux/amd64` images on the M3 Pro, push to `registry.mc02.dev`, pull on the server.**
+
+Rejected alternatives, with the reason each loses:
+
+| Option | Why not |
+|---|---|
+| **Build on the CX22** | `backend/Cargo.toml` sets `lto = true` + `codegen-units = 1` over **427 crates**. Fat LTO links the whole program in one single-threaded process; peak RSS is estimated at 2.5–4 GB against ~1.5–2.5 GB free with the stack running. OOM is the expected outcome, not a tail risk. And *rollback would also be a build* — 35–60 min under pressure. |
+| **CI (Gitea on Pi 5)** | Ruled out by you, and correct: a Pi 5 is strictly worse than the CX22 for a fat-LTO link. |
+| **True cross-compilation** (`--target x86_64-unknown-linux-musl`) | The dependency graph has **four C-compiling crates** — `ring` (per-arch assembly), `zstd-sys`, `libdeflate-sys`, `libsqlite3-sys` — so it needs a real musl cross-toolchain. Alpine has no such package for an arm64 host; the working answer is `cargo-zigbuild`, which means a new base image and a new toolchain days before an unrepeatable event. **Slow but certain beats fast but novel.** |
+
+Emulated build is the right call because the cost lands where it is free: on your laptop, days
+early, with nothing depending on it. Estimated wall-clock for the backend is **25–60 min with
+Rosetta enabled** (hours without it) — start it and walk away.
+
+> **Escape hatch if emulation is unbearable:** spin up a temporary Hetzner CPX41 (8 vCPU/16 GB) in
+> the same account, build natively, push, destroy it. Costs cents, same commands, no repo change.
+
+---
+
+## 2. Repo changes — ALREADY APPLIED
+
+> **Status: these are done, in the working tree.** They are documented here so you know what
+> changed and why, not as work to repeat. Run `git diff` to review before committing.
+
+### 2.1 `.dockerignore` files (were absent)
+
+`backend/.dockerignore`:
+```
+target/
+```
+
+`frontend/.dockerignore`:
+```
+node_modules/
+.svelte-kit/
+build/
+```
+
+Without these, the first time you run `cargo build` or `npm install` locally, every image build
+ships a multi-GB context to an *emulated* builder. Worse: `frontend/Dockerfile:9` does `COPY . .`
+**after** `npm ci`, so a macOS `node_modules/` would be merged over the container's Linux one.
+
+### 2.2 Switch compose from `build:` to `image:`
+
+In `docker-compose.yml`, replace the `build:` block on `app` and `frontend`:
+
+```yaml
+  app:
+    image: registry.mc02.dev/eventsnap/app:${EVENTSNAP_VERSION:?set EVENTSNAP_VERSION in .env}
+```
+```yaml
+  frontend:
+    image: registry.mc02.dev/eventsnap/frontend:${EVENTSNAP_VERSION:?set EVENTSNAP_VERSION in .env}
+```
+
+Two deliberate properties:
+- The `:?` form **fails loudly** on an unset variable instead of resolving to an empty tag.
+- **Removing `build:` entirely is a safety feature.** With no `build:` key on the server, a wrong
+  tag is an instant `manifest unknown` — never a surprise 45-minute compile on the production box.
+
+New `docker-compose.build.yml` (opt-in, Mac only — follows the convention stated in
+the header of `docker-compose.dev.yml` that overlays are never auto-loaded):
+
+```yaml
+# Build overlay. NOT loaded automatically. Used only where images are BUILT — never on the
+# production server, which pulls.
+#   docker compose -f docker-compose.yml -f docker-compose.build.yml build
+services:
+  app:
+    build: { context: ./backend, dockerfile: Dockerfile }
+  frontend:
+    build: { context: ./frontend, dockerfile: Dockerfile }
+```
+
+`e2e/docker-compose.test.yml` keeps its own `build:` blocks, so the e2e gate is unaffected.
+
+### 2.3 Add log rotation to every service
+
+`docker-compose.yml` currently sets **no logging config**, so the default `json-file` driver keeps
+logs forever, on the same filesystem as Postgres and the media. Add to each of the four services:
+
+```yaml
+    logging:
+      driver: json-file
+      options: { max-size: "10m", max-file: "3" }
+```
+
+(Equivalently, set it once in `/etc/docker/daemon.json` — but that restarts the Docker daemon,
+so do it *before* the stack is live, not after.)
+
+---
+
+## 3. `.env` — production values
+
+```bash
+# ── Identity ──────────────────────────────────────────────────────────────
+DOMAIN=<your domain>
+EVENT_NAME=<...>
+EVENT_SLUG=<...>
+
+# ── Image version (NEW — drives the image: tags in docker-compose.yml) ────
+EVENTSNAP_VERSION=v0.13.0
+
+# ── Secrets — ALL of them, before the first `up -d` ───────────────────────
+JWT_SECRET=<openssl rand -hex 64>
+POSTGRES_PASSWORD=<openssl rand -hex 24>
+DATABASE_URL=postgres://eventsnap:<SAME PASSWORD>@db:5432/eventsnap
+ADMIN_PASSWORD_HASH='<docker run --rm caddy:2-alpine caddy hash-password --plaintext "pw">'
+
+# ── Paths — must match the volume mounts ──────────────────────────────────
+MEDIA_PATH=/media          # pinned by compose anyway, but keep consistent
+EXPORT_PATH=/exports       # NOT pinned by compose — see the trap in §7.3
+
+# ── Sizing (see the two corrections below) ────────────────────────────────
+DATABASE_MAX_CONNECTIONS=30
+COMPRESSION_WORKER_CONCURRENCY=2
+
+# ── Comments off, likes + captions on ─────────────────────────────────────
+COMMENTS_ENABLED=false
+
+# ── Logging (NEW — production currently defaults to DEBUG) ────────────────
+RUST_LOG=eventsnap_backend=info,tower_http=warn
+```
+
+### Two corrections to the repo's own advice — do not follow `.env.example` here
+
+**`COMPRESSION_WORKER_CONCURRENCY`: keep `2`. Do NOT raise to 4, and do NOT raise the app memory
+limit to 2G.** `.env.example:92-98` justifies both on the premise that "each worker can run an
+ffmpeg transcode". **There is no transcode anywhere in this codebase.** `services/video.rs::run_ffmpeg`
+runs `ffmpeg -ss <t> -i <src> -vframes 1 -vf scale=…` — a single poster frame. Video originals are
+stored and served byte-for-byte.
+
+The real memory consumer is the **image** path: `image` 0.25's `resize` builds an `Rgba32F`
+intermediate at **16 bytes/px**, sized `source_width × target_height`, which the 256 MiB decode
+guard in `imaging::decode_limits` does not cover. Estimated peak per photo:
+
+| Source | Peak (decode + display resize) |
+|---|---|
+| 12 MP (typical phone) | ~145 MB |
+| 24 MP (iPhone Pro default) | ~223 MB |
+| 48 MP ("Max" mode) | ~354 MB |
+
+At concurrency 2, two 48 MP photos ≈ 800 MB against the 1 GiB cap — ~25% margin. At concurrency 4
+the same pair is ~1.5 GB → **OOM**. And app=2G + db=1G + 256M + 256M + ~370 MB OS/Docker ≈ 3954 MiB
+against ~3910 MiB MemTotal — the box is oversubscribed before a single photo arrives.
+
+**`quota_tolerance`: keep `0.75`. Raising it does not make anything more generous for a real guest.**
+See §4.
+
+---
+
+## 4. Admin dashboard configuration
+
+These live in the DB `config` table, not `.env`. Changes take effect on the **next request** —
+`patch_config` invalidates the cache synchronously (`admin::patch_config`). No restart needed. Set them
+after the first deploy, before the event.
+
+| Key | Default | **Set to** | Why |
+|---|---|---|---|
+| `upload_rate_per_hour` | 100 | **1000** | A guest multi-selecting 100 photos hits exactly 100. Client backs off and resumes, but this makes it a non-issue. |
+| `feed_rate_per_min` | 60 | **240** | Headroom for reconnect bursts. |
+| `social_rate_per_min` | 120 | **600** | This is the **likes** limiter — the interaction you're keeping. |
+| `export_rate_per_day` | 3 | **20** | **One shared bucket across both archives** — `enforce_export_rate` keys on `export:{user_id}` regardless of which archive is being fetched. Downloading Gallery.zip + Memories.zip costs 2 of 3; one retry locks a guest out for 24 h. The limit is now charged when the download *ticket* is minted, so being over it produces a visible German error instead of a tap that silently does nothing. |
+| `max_video_size_mb` | 500 | **leave at 500** | See below. |
+| `quota_tolerance` | 0.75 | **leave at 0.75** | See below. |
+| `quota_enabled`, `storage_quota_enabled`, `rate_limits_enabled` | true | **leave on** | This is the only disk-full safety net. |
+
+**Ignore `estimated_guest_count` and `upload_count_quota_enabled`.** Both are seeded, validated and
+rendered in the admin UI — and **read by no code at all** (verified by grep across `backend/src`).
+Changing them does nothing. `estimated_guest_count` in particular does *not* feed the quota formula.
+
+### Why quotas are already as generous as you want
+
+```
+per_user_limit = floor(free_disk × quota_tolerance / max(active_uploaders, 1))
+```
+`active_uploaders` is `SELECT COUNT(DISTINCT user_id) FROM upload WHERE deleted_at IS NULL`
+(`upload::compute_storage_quota`) — **people who actually uploaded**, not guests who joined. The 70 guests who never
+upload are not in the denominator; their share flows to the photographers automatically. **The
+redistribution you asked for is already the design.**
+
+With ~28 GB free and a realistic 30 people actually uploading, each gets **~700 MB** — against an
+expected ~1.25 GB for the *entire event*. Nobody will be blocked.
+
+Raising `quota_tolerance` would only raise the **saturation ceiling** (media converges to
+`t/(1+t)` of free space: 43% at 0.75, 50% at 1.0). It does nothing for a real guest at your volume,
+and it eats the headroom the keepsake needs — the export preflight reserves `media × 1.10 × 2`
+because `Gallery.zip` and `Memories.zip` each store media **uncompressed** (`export.rs` — both archives store media uncompressed).
+
+**Why `max_video_size_mb` stays at 500.** An earlier draft of this runbook said to lower it to 250,
+on the grounds that there is no client-side size check. That was wrong on both halves:
+
+- A client guard **does** exist — `frontend/src/routes/upload/+page.svelte` rejects anything over
+  `HARD_MAX_UPLOAD_BYTES` (576 MiB) before a byte leaves the phone, alongside the HEIC reject.
+- That guard is a **compile-time constant**, not the DB value. Lowering `max_video_size_mb` therefore
+  changes nothing on the client: the picker still accepts the clip, the phone still starts sending
+  it, and the *server* aborts it partway (`stream_field_to_file` does stop at the cap, so the AP is
+  not saturated for the full file — but the guest still gets a mid-upload failure where 500 MB would
+  simply have worked).
+
+So lowering it is a pure capacity decision with no UX upside, and at your volume there is no capacity
+problem to solve. Leave it. If you ever do want a smaller ceiling to bite on the phone, the constant
+above has to move with it.
+
+---
+
+## 5. Pre-flight — T‑7 days (do NOT leave this to event week)
+
+### Registry
+
+From **both** the Mac and the server, as **the user who will deploy** (root's `docker login` does
+not help a non-root deploy — credentials go to `~/.docker/config.json`):
+
+```bash
+getent hosts registry.mc02.dev                 # DNS resolves from the server
+curl -fsSI https://registry.mc02.dev/v2/       # TLS must be PUBLICLY trusted; Docker rejects
+                                               # self-signed without extra daemon config
+docker login registry.mc02.dev
+```
+
+Also confirm: the `eventsnap` repository/namespace exists if your registry requires pre-creation
+(Harbor does; plain `registry:2` does not), and the registry host has ~500 MB free per release.
+
+### DNS and TLS — get this wrong and you are locked out for an hour
+
+The DNS **A record must point at the CX22 before the first `docker compose up -d`**, because Caddy
+attempts certificate issuance on boot. Let's Encrypt allows only **5 failed validations per hostname
+per hour** (refilling 1 per 12 min). A misconfigured DNS record plus a few impatient restarts will
+rate-limit you out of getting a certificate at all.
+
+- Deploy days early so issuance happens with no time pressure.
+- **Never delete the `caddy_data` volume** — it holds the certificate and the ACME account key.
+- Never run `docker compose down -v`. It destroys `postgres_data`, `media_data`, `exports_data`
+  *and* `caddy_data`.
+
+### Host preparation
+
+```bash
+docker compose version          # must be v2.x — the deploy.resources limits need it
+free -h && swapon --show        # Hetzner images ship no swap
+df -h /var/lib/docker           # want ≥ 25 GB free
+```
+
+**Add 2 GB of swap** as an OOM cushion — a compression spike that would otherwise kill the container
+instead swaps out cold pages and merely runs slowly:
+
+```bash
+fallocate -l 2G /swapfile && chmod 600 /swapfile && mkswap /swapfile && swapon /swapfile
+echo '/swapfile none swap sw 0 0' >> /etc/fstab
+sysctl -w vm.swappiness=10 && echo 'vm.swappiness=10' > /etc/sysctl.d/99-swap.conf
+```
+
+> **Gotcha:** Compose sets each container's `Memory` limit but leaves `MemorySwap` unset, and Docker
+> then allows swap equal to the memory limit — so adding host swap silently **doubles** every
+> container ceiling. If you add swap, also add `memswap_limit: 1152m` to `app` and `db`, and
+> `memswap_limit: 320m` to `frontend` and `caddy` (service-level, not under `deploy:`).
+
+---
+
+## 6. Build and push — on the Mac
+
+Enable **Docker Desktop → Settings → General → "Use Rosetta for x86_64/amd64 emulation"**, and set
+**Resources → Memory ≥ 8 GB** (the fat-LTO step will OOM inside the VM otherwise, even with 18 GB on
+the host).
+
+```bash
+docker run --rm --platform linux/amd64 alpine uname -m     # must print x86_64
+
+cd /Users/fabianhammprivat/Projects/EventSnap
+VERSION=v0.13.0                          # latest existing tag is v0.12.0 — see §9 before reusing it
+ROLLBACK=v0.13.0-a                       # the SAME source, tagged twice; §9 explains why
+SHA=$(git rev-parse --short HEAD)
+
+docker buildx create --name eventsnap --use 2>/dev/null || docker buildx use eventsnap
+
+docker buildx build --platform linux/amd64 \
+  -t registry.mc02.dev/eventsnap/app:$VERSION \
+  -t registry.mc02.dev/eventsnap/app:$ROLLBACK \
+  -t registry.mc02.dev/eventsnap/app:$SHA \
+  --push ./backend
+
+docker buildx build --platform linux/amd64 \
+  -t registry.mc02.dev/eventsnap/frontend:$VERSION \
+  -t registry.mc02.dev/eventsnap/frontend:$ROLLBACK \
+  -t registry.mc02.dev/eventsnap/frontend:$SHA \
+  --push ./frontend
+```
+
+### Verify the architecture — never skip this
+
+An arm64 image pulls fine and then dies with `exec format error`. Catch it here, not on the server:
+
+```bash
+docker buildx imagetools inspect registry.mc02.dev/eventsnap/app:$VERSION
+docker buildx imagetools inspect registry.mc02.dev/eventsnap/frontend:$VERSION
+# Both MUST report  Platform: linux/amd64
+```
+
+Then prove the binary actually executes. `AppConfig::from_env()` runs before any DB connection
+(`main.rs` builds `AppConfig` before touching the pool), so this needs no database:
+
+```bash
+docker run --rm --platform linux/amd64 \
+  -e APP_ENV=production -e JWT_SECRET=x -e DATABASE_URL=x -e EVENT_SLUG=x \
+  registry.mc02.dev/eventsnap/app:$VERSION
+```
+Expect the "Refusing to start in production … placeholder" message. **That message means the amd64
+binary ran.** `exec format error` means the architecture is wrong.
+
+**Tag policy: never `latest` in production.** With `latest` you cannot tell what is running,
+rollback becomes a registry re-push (impossible if the registry is down — exactly when you need it),
+and `restart: unless-stopped` after a reboot is ambiguous. Two immutable tags per build: semver and
+short SHA. Deploy by semver.
+
+---
+
+## 7. First deploy — on the server
+
+```bash
+# Directory name matters: volumes are prefixed with it, and README's backup commands
+# hardcode the `eventsnap_` prefix.
+git clone <repo> eventsnap && cd eventsnap
+cp .env.example .env && nano .env          # every value from §3
+docker login registry.mc02.dev
+docker compose pull                        # must fully succeed before anything starts
+```
+
+### 7.1 The secret pre-flight — run this before the first `up -d`, always
+
+```bash
+docker compose run --rm --no-deps app
+```
+
+`--no-deps` means `db` never starts, so **no Postgres data directory is initialised**.
+`AppConfig::from_env()` runs first and reports *every* placeholder at once (`config::validate_secrets`).
+When the only remaining complaint is a database *connection* failure, the secrets are good.
+
+**Why this step exists.** `POSTGRES_PASSWORD` is applied **only at initdb**. `docker-compose.yml`
+starts `db` in the same command as `app`, so a single `up -d` with a placeholder bakes the wrong
+password in permanently — the app then loops on `password authentication failed`, and the only exits
+are `ALTER ROLE` or `down -v`, which deletes the database, the media and the exports. The repo
+describes this trap at `backend/src/db.rs` (`explain_auth_failure`); this command is what avoids it.
+
+### 7.2 Bring it up
+
+```bash
+# `.env` is consumed by docker compose, not by your shell — load it before using $DOMAIN.
+set -a; . ./.env; set +a
+
+docker compose up -d
+docker compose logs -f app          # wait for "database connected and migrations applied"
+curl -fsS https://$DOMAIN/health    # → ok  (503 means the app is up but the DB is not)
+```
+
+### 7.3 Verify the three things compose does *not* pin
+
+`MEDIA_PATH` is pinned to `/media` on the `app` service in `docker-compose.yml`. Its siblings
+are not:
+
+```bash
+docker compose exec app printenv DATABASE_URL EXPORT_PATH ADMIN_PASSWORD_HASH
+```
+
+1. **`DATABASE_URL`** must contain `@db:5432`. A dev `.env` points it at `@localhost`, which inside
+   the container is the app itself.
+2. **`EXPORT_PATH`** must be `/exports`. Anywhere else and the keepsake archives are written to the
+   container's writable layer and **vanish on the next `up -d`** — including on a rollback.
+3. **`ADMIN_PASSWORD_HASH`** must match `.env` **byte for byte.**
+
+**Then actually log in to `/admin` with the real password.** This is not optional politeness:
+
+- The production secret guard only rejects *placeholders* (`config::looks_placeholder`). A hash **corrupted**
+  by shell or Compose escaping is not a placeholder — the app boots green, `/health` says `ok`, and
+  every admin login 401s.
+- The Admin user row is created **by a successful admin login** (`auth::handlers::admin_login`), and
+  promoting anyone to Host requires an Admin or Host (`auth::middleware`'s role guard). **No admin login
+  ⇒ no host, ever** ⇒ you cannot close the event, release the gallery, ban anyone, reset a PIN, or
+  change any limit — for the whole event.
+
+Per the Compose spec, single-quoted `env_file` values *are* used literally and the quotes are
+stripped, so the single-quote form in `.env.example` is correct. The comment in
+`docker-compose.dev.yml` claiming production has the same bug is **stale**. Verify anyway —
+the cost of checking is 10 seconds; the cost of being wrong is the whole event.
+
+**Promote a second person to Host** once you are in, so a single lost session is not fatal.
+
+---
+
+## 8. Post-deploy verification
+
+```bash
+set -a; . ./.env; set +a                                     # $DOMAIN comes from .env, not your shell
+
+docker compose ps                                            # all four healthy
+docker inspect -f '{{.HostConfig.Memory}}' eventsnap-app-1   # must be 1073741824, not 0
+curl -fsS https://$DOMAIN/health                             # ok — now a real DB check, not a constant
+docker compose exec app printenv COMMENTS_ENABLED RUST_LOG
+docker builder prune -af && docker image prune -f            # reclaim build cache
+df -h /var/lib/docker
+```
+
+Then, from a phone on cellular (not the office wifi):
+
+- [ ] Join as a guest with a PIN
+- [ ] Upload a photo → appears in the feed within a few seconds
+- [ ] Upload a video → plays back (iOS Safari range requests)
+- [ ] Add a caption, add a like — **no comment UI anywhere**
+- [ ] Admin login works; host dashboard reachable
+- [ ] Release the gallery on a test event and download both archives
+
+---
+
+## 9. Rollback
+
+### There is no older image you can roll back to. Build the rollback target yourself.
+
+Read this before the event, not during it. The obvious move — drop `EVENTSNAP_VERSION` back to the
+previous released tag — **takes the app down permanently** and looks like a crash loop with no
+explanation:
+
+```
+$ git ls-tree --name-only v0.12.0 backend/migrations/ | wc -l
+12          # 6 migrations. HEAD has 22.
+$ git rev-list --count v0.12.0..HEAD
+154
+```
+
+`db.rs` runs `sqlx::migrate!()` with no `set_ignore_missing`, so an image built from a 6-migration
+tree, booting against a database that already carries versions 007–022, returns `VersionMissing`.
+`create_pool` errors, `main` exits 1, and `restart: unless-stopped` restarts it forever — with Caddy
+still routing traffic to it. (`014_export_epoch.up.sql` documents this failure mode; §0 restates it.)
+No `v0.12.0` image was ever built or pushed either, so the pre-pull would fail with
+`manifest unknown` before you ever got that far.
+
+**So: at build time, tag the SAME frozen commit twice.** Two identical images, two names. The
+rollback then swaps to a binary that is bit-for-bit what you tested and carries the identical
+migration set, which makes it a genuine no-op rather than a gamble:
+
+```bash
+# In §6, push both tags from the one build:
+VERSION=v0.13.0
+ROLLBACK=v0.13.0-a          # same source, different name — the rollback target
+
+docker buildx build --platform linux/amd64 \
+  -t registry.mc02.dev/eventsnap/app:$VERSION \
+  -t registry.mc02.dev/eventsnap/app:$ROLLBACK \
+  --push ./backend
+# ...and the same two tags for ./frontend
+```
+
+**Rolling back — ~30 seconds, no network:**
+
+```bash
+sed -i 's/^EVENTSNAP_VERSION=.*/EVENTSNAP_VERSION=v0.13.0-a/' .env
+docker compose up -d app frontend
+```
+
+This only works offline if both are already resident. **Pre-pull all four at T‑2:**
+
+```bash
+docker pull registry.mc02.dev/eventsnap/app:v0.13.0
+docker pull registry.mc02.dev/eventsnap/frontend:v0.13.0
+docker pull registry.mc02.dev/eventsnap/app:v0.13.0-a
+docker pull registry.mc02.dev/eventsnap/frontend:v0.13.0-a
+docker image ls | grep eventsnap        # confirm all four
+```
+
+Once resident, `up -d`, reboots, restarts and rollbacks need **zero** registry contact. That single
+step makes a registry outage on event day irrelevant.
+
+Be clear-eyed about what this buys you: an identical image cannot undo a bad *release*, only an
+image that got corrupted or a container that wedged — and `docker compose restart app` already
+covers both. It exists so that the rollback line in the emergency card is safe to run rather than
+catastrophic. **If you genuinely need to undo a code change during the event, you cannot; freeze
+early enough that you never have to.**
+
+**Across a migration boundary — avoid by freezing.** If you must: all 22 migrations have paired
+`.down.sql` files, but **none of them removes its own `_sqlx_migrations` row**, so that second step
+is mandatory and undocumented:
+
+```bash
+docker compose stop app
+# `sh -c` so the CONTAINER expands the credentials. They live in the container's environment
+# and in .env — not in your shell — so an unwrapped `-U "$POSTGRES_USER"` sends `-U ""` and
+# psql answers `FATAL: role "" does not exist`.
+docker compose exec -T db sh -c \
+  'psql -U "$POSTGRES_USER" -d "$POSTGRES_DB" -v ON_ERROR_STOP=1' \
+  < backend/migrations/0NN_x.down.sql
+docker compose exec -T db sh -c \
+  'psql -U "$POSTGRES_USER" -d "$POSTGRES_DB" -c "DELETE FROM _sqlx_migrations WHERE version = NN;"'
+```
+
+Migration **014** is the only destructive one on the way up, and the only one with a rehearsal
+harness — `backend/scripts/rehearse-014.sh`. Run it once against a real dump before the event.
+
+**Registry-down transport fallback** (layers are already compressed — do not add gzip):
+
+```bash
+docker save registry.mc02.dev/eventsnap/app:v0.13.0 \
+            registry.mc02.dev/eventsnap/frontend:v0.13.0 | ssh root@SERVER 'docker load'
+```
+
+---
+
+## 10. Backup
+
+Full commands are in `README.md:315-435` and are correct — `pg_dump --clean --if-exists`, plus
+`alpine tar` out of `eventsnap_media_data` and `eventsnap_exports_data`, mounted at `/src` (not
+`/media`), with `chown -R 100:101` on restore because the app runs non-root and BusyBox tar has no
+`--same-owner`. There is deliberately no script.
+
+Two gaps the README does not cover:
+
+1. **Nothing backs up `.env`**, which holds the only copy of `POSTGRES_PASSWORD`. A dump you cannot
+   authenticate against is not a backup. Copy `.env` off the box, encrypted, once it is final.
+2. **Timing.** Do not use nightly cron — every irreplaceable byte is created inside one evening.
+   Take the dump and the media tarball back-to-back **the night of the event, after locking uploads
+   from the host dashboard**, so the pair is consistent.
+
+---
+
+## 11. Disk — the numbers for 40 GB
+
+Expected event (100 photos @ ~4 MB, 5 videos @ ~150 MB). Originals are **always kept** and never
+lossily recompressed; derivatives are a ≤800 px preview and a ≤2048 px display JPEG.
+
+```
+media (originals + derivatives)                    ~1.25 GB
+peak during keepsake build (+ Gallery + Memories)  ~3.5 GB
+OS + Docker images + Postgres + logs               ~4.5 GB
+────────────────────────────────────────────────────────────
+total                             ~11–13 GB of ~36 GB usable
+```
+
+**40 GB fits with roughly 3× headroom**, provided you build elsewhere (a server-side build adds
+3–5 GB of cache that permanently shrinks the guest quota, because the quota is recomputed against
+*live* free space on every upload).
+
+The `README.md:295-299` "ENOSPC" projection models guests **saturating the quota** (~12 GB of
+media), not 100 photos. That scenario needs ~10× your expected volume — and it degrades gracefully:
+the export preflight refuses up front rather than hitting ENOSPC mid-write, and the host dashboard
+warns when free space drops below 10 GB or below the keepsake requirement (`handlers::host`'s low-disk thresholds).
+
+---
+
+## 12. Known issues you are shipping with
+
+None of these has a fix in this runbook; they are listed so nothing is a surprise. Severity is
+scored purely by "would this interrupt you during the event".
+
+### Fixed in this pass
+
+| Was | Fix |
+|---|---|
+| **Queued uploads never rehydrated** — `loadQueue()` had one call site, so a guest whose PWA was evicted mid-upload and reopened onto `/feed` had pending photos in IndexedDB that nothing ever read. Silent photo loss. | `loadQueue()` now runs on every authenticated boot in `+layout.svelte`. |
+| **A corrupted `ADMIN_PASSWORD_HASH` booted green** and only failed at admin login — unrecoverable mid-event. | `config.rs` now validates the bcrypt *shape*, not just placeholder-ness, and refuses to start. |
+| **SSE reconnect thundering herd** — flat 500 ms jitter regardless of backoff. | Jitter now scales with the delay; the feed's in-place refresh is spread over 800–2800 ms. |
+| **No client-side size/HEIC pre-check** — a doomed 600 MB upload saturated the venue AP before being rejected. | Certain-rejects are refused before any bytes leave the phone. |
+| **Upload-queue UI unreachable** — the badged FAB opened the picker, not the queue, so a failed upload had no retry button. | The upload sheet now shows a queue entry whenever the badge is non-zero. |
+| **A mid-event 401 left a dead screen** with no nav and no URL bar. | `api.ts` now returns the guest to `/join` (skipping the auth routes themselves). |
+| **Rate limiter could panic while holding its mutex** (`timestamps[0]` with `max == 0`), poisoning it process-wide. | Uses `first()` with a fail-open guard. |
+| **Hashtag deadlock** — `edit_upload` and `add_comment` upserted tags in client/text order. | Both now sort + dedup on the normalised key, matching the upload path. |
+| **Unbounded Docker logs + debug-level app logging.** | `logging:` caps every service at 30 MB; `RUST_LOG` documented in `.env.example`. |
+
+### Fixed in the readiness pass — behaviour you should know about
+
+These change what you will observe on the night, so they are listed separately from the table above.
+
+| Was | Now |
+|---|---|
+| **Any ffmpeg-level failure DESTROYED the video.** A poster-frame error — ffmpeg hanging on a truncated `.mov`, an ENOSPC on `thumbnails/`, a DB blip — propagated into the give-up path, which soft-deleted the upload. Reproduced live: on a host with no ffmpeg the clip was gone ~6 s after its `201 Created`. | No failure in the video branch can fail the upload. The clip stays in the feed and plays from its original; only the poster is missing. |
+| **A lost response duplicated the photo.** Every retry minted a fresh upload id, so a phone that lost the reply and re-sent — manually, or automatically on reconnect — stored the same photo two or three times and paid quota for each. | The client sends a `client_upload_id`; migration 022 makes it unique. A retry returns `200` with the original row. Verified live: three sends → one row, quota charged once. |
+| **`/health` returned a constant `"ok"`.** The container reported healthy while every request 500'd. | Runs `SELECT 1` with a 2 s timeout: `200 ok` / `503`. Verified live: 200 → stop Postgres → 503 → start Postgres → 200, **with no app restart** (the pool revalidates on acquire). Note that Compose does not restart on an unhealthy probe — this is a diagnostic, deliberately not wired to automatic recovery, because a restart would truncate in-flight uploads to "fix" an outage that clears on its own. |
+| **Abandoned `.tmp` uploads were never reclaimed** (see the old issue #2 below — it is now fixed, not deferred). | Swept at boot and hourly, keyed on modification time so a live upload can never age into it. |
+| **A full disk made itself worse.** ENOSPC was retried three times, then refunded the quota, soft-deleted the row and *kept* the bytes — freeing nothing and inviting an immediate re-upload into the same full disk. | ENOSPC is classified separately: no retry, no refund, no delete. The row stays live and the photo is served from its original until there is room to compress it. |
+| **The keepsake download failed invisibly.** The rate limit was enforced inside the iframe navigation, so an over-limit guest tapped and *nothing happened*, forever. | Charged when the ticket is minted — a normal `fetch` — so it surfaces as a German message naming the daily window. Raise `export_rate_per_day` per §4 anyway. |
+| **`/host` and `/admin` subscribed to SSE but never opened the connection**, so the keepsake progress bar froze after a release. | Both connect on mount and disconnect on destroy, as `/export` already did. |
+| **`?limit=-5` on the feed returned a 500.** | Clamped at both ends. |
+| **All recurring hygiene lived in one unsupervised task** — one panic silently stopped session pruning, media reclamation and the temp sweep for the rest of the event. | Supervised and re-spawned, with an error log. |
+| **No pool acquire or statement timeout.** A DB blip parked every request for 30 s. | 5 s acquire, `statement_timeout=15s`, `lock_timeout=5s`. |
+
+### Still open — accepted for this event
+
+| # | Issue | Impact |
+|---|---|---|
+| 1 | **Guest delete/caption-edit after release invalidates both keepsake archives** (`upload.rs`), forcing a rebuild with a 20 s debounce. **No-op before release**, so it cannot bite during the event. | Post-event only. Left alone because blocking guest deletes has real privacy downsides — that is a product call, not a bug fix. |
+| 2 | **Video poster frames do not regenerate after a restart** mid-compression (the backfill filters `mime_type LIKE 'image/%'`). | Cosmetic — the video still plays; only its poster is missing. |
+| 3 | **SSE keep-alives are sent as SSE comments** (`:ping`), which the browser's EventSource parser discards without dispatching. A client therefore cannot implement a pure silence timer to detect a half-open socket. | Worked around client-side: the feed runs a jittered 60–120 s `/feed/delta` backstop and reconnects when a poll returns content the stream never delivered. A cleaner fix is to emit keep-alives as a *named* event; that is a coordinated backend+frontend change, not worth making during a freeze. |
+| 4 | **The lightbox stops at the end of the loaded page** — stepping past the last loaded photo does not fetch the next one. | The guest scrolls the feed (which does page) and re-opens. |
+
+---
+
+## 13. Event-day emergency card
+
+Assume you have a phone and two minutes.
+
+```bash
+cd ~/eventsnap
+
+# FIRST LINE, ALWAYS. `.env` is read by docker compose, NOT by your shell — without this,
+# every `$DOMAIN` below expands to nothing and `curl https:///health` reads like an outage
+# when the site is fine.
+set -a; . ./.env; set +a
+
+# Is it alive? (200 = app AND database are answering; 503 = the app is up, the DB is not)
+curl -fsS https://$DOMAIN/health
+
+# What is broken?
+docker compose ps
+docker compose logs --tail=100 app
+
+# Nuclear option that is SAFE (keeps all data):
+docker compose restart app
+
+# Roll back to the identically-built sibling image — see §9 for what this can and cannot fix.
+# Do NOT substitute an older release tag here; it will crash-loop on the migration set.
+sed -i 's/^EVENTSNAP_VERSION=.*/EVENTSNAP_VERSION=v0.13.0-a/' .env && docker compose up -d app frontend
+
+# Disk check
+df -h /var/lib/docker
+```
+
+**NEVER** run `docker compose down -v`. It deletes the database, all media, all exports and the TLS
+certificate. There is no undo.
+
+Most limits are changeable from the **admin dashboard without a restart** — reach for that before
+touching the shell.

PROJECT.md

@@ -709,7 +709,7 @@ CREATE TABLE config (
 INSERT INTO config (key, value) VALUES
     ('max_image_size_mb',       '20'),
     ('max_video_size_mb',       '500'),
-    ('upload_rate_per_hour',    '10'),
+    ('upload_rate_per_hour',    '100'),  -- raised from 10 in migration 015 (guests upload bursts of 10-20)
     ('feed_rate_per_min',       '60'),
     ('export_rate_per_day',     '3'),
     ('quota_tolerance',         '0.75'),
@@ -1133,16 +1133,19 @@ eventsnap/
 
 ### Backup Strategy
 
-```bash
-# Daily (e.g. as a separate Compose service or cron on the VPS)
-pg_dump $DATABASE_URL | gzip > /media/backups/db_$(date +%Y-%m-%d).sql.gz
+Three artefacts in three places: the database, the `media_data` volume
+(originals + derivatives), and the **separate** `exports_data` volume. See
+[README.md](README.md#backup) for the exact commands.
 
-# Weekly: rsync /media volume to Hetzner Storage Box
-rsync -az /opt/eventsnap/media/ \
-  user@u123456.your-storagebox.de:backup/eventsnap/
-```
+Everything runs through `docker compose` / `docker run`, because `DATABASE_URL`
+and the `/media` and `/exports` paths only exist inside the compose network —
+they are not host paths, and `DATABASE_URL` is never exported into an operator's
+shell.
 
-The `/media` volume contains originals, previews, thumbnails, generated exports, and DB backups — a single volume to back up.
+Export archives are deliberately outside `MEDIA_PATH` (`EXPORT_PATH=/exports`): a
+keepsake contains every photo in the event, and keeping it off the media tree is
+what stops it being reachable except through the ticket-gated handler. A backup
+of the media volume alone silently loses every generated keepsake.
 
 ---
 

README.md

@@ -34,7 +34,6 @@ A guest scans the QR code on their way in, types their name, and is immediately
 ### Planned (v1.x)
 
 - Individual file download button
-- Low-disk alert (< 10 GB free)
 - Event banner / cover image
 - Chunked resumable upload for large videos
 - Host-curated story highlights
@@ -98,33 +97,135 @@ eventsnap/
 git clone https://git.mc02.dev/fabi/EventSnap.git eventsnap
 cd eventsnap
 
-# 2. Configure environment
+# 2. Configure environment — set EVERY secret NOW, before step 3.
 cp .env.example .env
-nano .env   # set DOMAIN, JWT_SECRET, ADMIN_PASSWORD_HASH, EVENT_NAME, etc.
+nano .env   # DOMAIN, EVENT_NAME, EVENT_SLUG,
+            # JWT_SECRET, ADMIN_PASSWORD_HASH,
+            # POSTGRES_PASSWORD *and* the same password inside DATABASE_URL
+            # (see "Generate required secrets" below)
 
 # 3. Start the stack
 docker compose up -d
 ```
 
+> **Set every secret before step 3 — `POSTGRES_PASSWORD` especially.** Postgres reads it **only
+> when it initialises its data directory**, which happens on the very first `docker compose up -d`.
+> Changing it in `.env` afterwards does not change the stored password: the app then authenticates
+> with the new one against a volume holding the old one, and you get a permanent restart loop with
+> `password authentication failed for user "eventsnap"`. The only fixes are restoring the old
+> password or `docker compose down -v`, which **deletes the database, the media and the exports**.
+> Getting it right once, up front, costs nothing; getting it wrong costs the volume.
+
 Caddy automatically obtains a Let's Encrypt certificate on first start. The app is live at `https://DOMAIN` within ~30 seconds.
 
-> **If the site never comes up:** with `APP_ENV=production` the backend **refuses to boot** while `JWT_SECRET`/`ADMIN_PASSWORD_HASH` still hold the `.env.example` placeholders (this is deliberate — a publicly-known signing key is worse than downtime). Caddy then waits on the unhealthy `app` container and never serves. Check `docker compose logs app` — a "Refusing to start … placeholder …" line means you skipped step 2. Rotate the secrets (see below) and restart.
+> **If the site never comes up:** with `APP_ENV=production` the backend **refuses to boot** while
+> `JWT_SECRET`, `ADMIN_PASSWORD_HASH` or the password inside `DATABASE_URL` still hold the
+> `.env.example` placeholders (this is deliberate — a publicly-known signing key or database
+> password is worse than downtime). Caddy then waits on the unhealthy `app` container and never
+> serves. Check `docker compose logs app` — a "Refusing to start … placeholder …" line lists
+> **every** unset secret at once, so one edit fixes them all.
+>
+> **If it comes up but keeps restarting with `password authentication failed for user
+> "eventsnap"`:** `POSTGRES_PASSWORD` was changed after the database volume was created. Postgres
+> applies that variable only at initialisation, so `.env` and the stored password have drifted
+> apart permanently. `docker compose logs app` spells this out. Before the event, with nothing
+> worth keeping:
+>
+> ```bash
+> docker compose down -v && docker compose up -d   # -v DELETES db + media + exports. No undo.
+> ```
+>
+> **Once the event has real data, never do that.** Put the original password back into
+> `DATABASE_URL`, or change the stored one instead:
+>
+> ```bash
+> docker compose exec db psql -U "$POSTGRES_USER" -c \
+>   "ALTER ROLE eventsnap WITH PASSWORD 'the-password-now-in-your-.env';"
+> ```
 
 > **Production note:** `docker compose up -d` does **not** expose the database — Postgres is reachable only on the internal Docker network. For local development where you need host access to Postgres, opt into the dev overlay explicitly:
 > ```bash
 > docker compose -f docker-compose.yml -f docker-compose.dev.yml up
 > ```
 
+### Updating an existing deployment
+
+> **`docker compose up -d` alone will NOT deploy your changes.** `app` and `frontend` are
+> `build:` services with no published image tag, and Compose has no source-change detection:
+> if an image with that name already exists it is reused. After a `git pull` the command
+> reports `Container … Running`, changes nothing, and **exits 0** — so a deploy that shipped
+> nothing looks exactly like a successful one. `--build` is what makes it real.
+
+```bash
+cd /path/to/eventsnap
+
+# 1. Back up first — migrations run automatically on boot and are not reversible in place.
+#    (See "Backup" below; the database dump is the one that matters here.)
+
+# 2. Fetch the new code.
+git pull
+
+# 3. Rebuild and restart the application services. --build is NOT optional.
+docker compose up -d --build
+
+# 4. Apply any Caddyfile change. Step 3 does NOT do this — see the warning below.
+docker compose up -d --force-recreate caddy
+
+# 5. Confirm the app came back up. Anything other than "ok" means check the logs.
+curl -fsS https://DOMAIN/health && echo
+
+# 6. Confirm a NEW image was actually built. Note the IMAGE ID before you start and
+#    compare — it must have changed. (Ignore the CREATED column; it reports the base
+#    layer's age, not this build's.) An unchanged ID means step 3 ran without --build
+#    and you are still serving the old code.
+docker compose images app frontend
+```
+
+Migrations are applied by the backend on startup, so step 3 covers them. If `app` stays
+unhealthy afterwards, `docker compose logs app` will name the failing migration — and note
+that a migration applied by a *newer* build is not removed by checking out an older commit,
+so rolling back code without restoring the database snapshot from step 1 leaves the schema
+ahead of the binary and the app refusing to boot.
+
+> **Why step 4 exists.** `--build` only rebuilds services that have a `build:` section, and
+> `caddy` is a pinned upstream image. Compose decides whether to recreate a container from its
+> *config hash*, which covers the mount **specification** (`./Caddyfile:/etc/caddy/Caddyfile:ro`)
+> but **not the file's contents** — so a `git pull` that changes `./Caddyfile` produces no
+> delta, Compose reports `Running`, and Caddy keeps serving its old config indefinitely. Exit
+> code 0 throughout.
+>
+> That is not hypothetical: the fix that made the keepsake download work on iOS
+> (`137c4ee`) touched the Caddyfile and four e2e files and nothing else, so **all** of its
+> production effect lives in that one file. Without step 4 you deploy it, watch both image IDs
+> change, and iOS downloads stay broken.
+>
+> `--force-recreate` rather than `restart` or `caddy reload`: the bind mount is resolved to an
+> **inode** when the container is created, and `git pull` replaces the file instead of editing
+> it in place, so the container can still be bound to the old, now-unlinked inode. A restart
+> then re-reads the stale content. Recreating the container re-resolves the path.
+
+`db` is never touched, and recreating `caddy` does not disturb the `caddy_data` volume, so the
+TLS certificate and all data volumes survive.
+
 ### Generate required secrets
 
 ```bash
 # JWT secret (64 random bytes)
 openssl rand -hex 64
 
-# Admin password hash (bcrypt, cost 12)
-htpasswd -bnBC 12 "" yourpassword | tr -d ':\n'
+# Database password (goes in BOTH DATABASE_URL and POSTGRES_PASSWORD)
+openssl rand -hex 24
+
+# Admin password hash (bcrypt). Uses an image the stack already pulls, so it needs
+# nothing installed on the host — `htpasswd` lives in apache2-utils, which a stock
+# VPS does not have. Emits cost 14 rather than 12; that is fine (admin login is
+# rate-limited and hashed off the async runtime), and any $2a/$2b/$2y hash verifies.
+docker run --rm caddy:2-alpine caddy hash-password --plaintext 'yourpassword'
 ```
 
+Wrap the resulting hash in **single quotes** in `.env` — see the note there; a bcrypt
+hash is full of `$`, and both Compose and dotenvy would otherwise eat those segments.
+
 ### Environment Variables
 
 See [.env.example](.env.example) for the full list with descriptions and defaults. Key variables:
@@ -162,23 +263,190 @@ See [.env.example](.env.example) for the full list with descriptions and default
 └────────┘
 ```
 
-- `/api/*` and `/media/*` → Rust backend
+- `/api/*` → Rust backend
 - Everything else → SvelteKit frontend (`adapter-node`)
-- Named volumes: `postgres_data`, `media_data`, `caddy_data`
+- Named volumes: `postgres_data`, `media_data`, `exports_data`, `caddy_data`
+
+Media is **not** served as static files. Every image goes through a
+visibility-checked alias (`/api/v1/upload/{id}/{preview,display,thumbnail,original}`)
+so a host takedown or a ban actually revokes access to the bytes.
+
+---
+
+## Sizing the disk
+
+`postgres_data`, `media_data` and `exports_data` are all Docker named volumes under
+`/var/lib/docker/volumes`, so **they share one filesystem**. Filling it does not
+degrade one subsystem — Postgres stops being able to write and the whole event goes
+down.
+
+**`Gallery.zip` and `Memories.zip` are each roughly a second copy of every original.**
+Both write their media `Compression::Stored`, and `Memories.zip` streams the untouched
+original for every video and for every image at or under 5 MB. So a release wants room
+for **two more copies of the gallery** on top of the gallery itself — which is what
+`required_free_bytes` encodes as `media × 1.1 × 2`.
+
+The per-user quota does **not** bound this. It is a fairness mechanism that divides
+free space between guests, and since it carries a floor (`MIN_QUOTA_LIMIT_BYTES`, so a
+guest's allowance stops shrinking as the party fills up) the aggregate ceiling it used
+to imply is gone. What bounds the disk is the **global gate in the upload handler**,
+which refuses any upload that would leave too little room to build the keepsake:
+
+```
+free_after_upload  <  media_after × 1.1 × 2  +  DISK_RESERVE_BYTES   →   refused
+```
+
+Solving that for the gallery size gives the real ceiling. On the **40 GB box this runs
+on**, with ~5 GB for the OS, Docker images (the runbook pre-pulls the rollback tag too)
+and Postgres:
+
+| Volume | Usable after baseline | Media ceiling | Free at release |
+|---|---|---|---|
+| 40 GB | ~35 GB | **~8 GB** | ~27 GB → both archives fit |
+| 80 GB | ~70 GB | ~19 GB | ~51 GB → both archives fit |
+
+**Uploads therefore stop at roughly 8 GB of media on a 40 GB box, not when the disk is
+full.** That is deliberate. 1000 photos at ~3.5 MB is ~3.5 GB and fits comfortably;
+video is what consumes the budget, so lower `max_video_size_mb` (seeded at 500) if you
+expect a lot of it. Refusing the 1001st upload is a far better outcome than accepting it
+and discovering at 01:00 that the archive can never be built.
+
+Two ways to buy headroom:
+
+- **Provision ~3× your expected media** on one volume (media + two archives), or
+- **give `exports_data` its own volume** so a full export cannot reach Postgres, and
+  size that one at ~2× expected media.
+
+None of this is silent. The upload gate refuses with a German message naming the cause,
+the export preflight refuses up front with both numbers rather than hitting ENOSPC
+halfway through a multi-GB write, a rebuild only reclaims the superseded generation
+**after** the new one lands (so a failed rebuild can never leave you with no archive at
+all), and the host dashboard warns as soon as the keepsake would not fit — which is the
+only point at which anyone can still do something about it.
 
 ---
 
 ## Backup
 
-```bash
-# Database snapshot
-pg_dump $DATABASE_URL | gzip > /media/backups/db_$(date +%Y-%m-%d).sql.gz
+There are **three** things to back up, and they live in three different places.
+`DATABASE_URL` and the container paths (`/media`, `/exports`) are meaningful only
+*inside* the compose network — they are not host paths, and `DATABASE_URL` is
+never exported into an operator's shell — so every command below runs through
+`docker compose` from the repo directory.
 
-# Weekly offsite sync (Hetzner Storage Box or similar)
-rsync -az /opt/eventsnap/media/ user@storagebox.example.com:backup/eventsnap/
+```bash
+# 1. Database snapshot. Runs pg_dump inside the db container (the app image has no
+#    postgres client), reading credentials from the compose environment.
+#    --clean --if-exists makes the dump SELF-CLEANING: without it the restore below
+#    aborts on the first "already exists" against a database that has ever booted,
+#    which is every database you would actually want to restore over.
+mkdir -p ./backups
+docker compose exec -T db \
+  sh -c 'pg_dump --clean --if-exists -U "$POSTGRES_USER" "$POSTGRES_DB"' \
+  | gzip > ./backups/db_$(date +%Y-%m-%d).sql.gz
+
+# 2. Uploaded media (originals + derivatives) out of the named volume.
+#    NOTE the mountpoint is /src, not /media: if the volume is ever empty, Docker
+#    pre-populates a fresh mount from the image's own directory, and alpine ships a
+#    /media containing cdrom/floppy/usb. Mounting somewhere the image has nothing
+#    avoids silently tarring (and polluting the volume with) those.
+docker run --rm \
+  -v eventsnap_media_data:/src:ro -v "$PWD/backups":/backup \
+  alpine tar czf /backup/media_$(date +%Y-%m-%d).tar.gz -C /src .
+
+# 3. Export archives — a SEPARATE volume (see the security note below).
+docker run --rm \
+  -v eventsnap_exports_data:/src:ro -v "$PWD/backups":/backup \
+  alpine tar czf /backup/exports_$(date +%Y-%m-%d).tar.gz -C /src .
+
+# Offsite sync of the three artefacts above.
+rsync -az ./backups/ user@storagebox.example.com:backup/eventsnap/
 ```
 
-The `/media` volume holds originals, previews, thumbnails, exports, and DB backups — a single path to back up.
+Volume names are prefixed with the compose project name — `eventsnap_` if you run
+from a directory called `eventsnap`. Confirm yours with `docker volume ls`.
+
+> **Exports are deliberately NOT under `/media`.** They live on their own
+> `exports_data` volume (`EXPORT_PATH=/exports`) because a keepsake archive
+> contains every photo in the event; keeping it outside the media tree is what
+> stops it being reachable except through the ticket-gated download handler.
+> Backing up only the media volume therefore loses every generated keepsake.
+
+### When to run it
+
+**A nightly cron is the wrong shape for this app.** Every irreplaceable byte is
+created inside one eight-hour window, and nobody can retake a wedding. Run the three
+commands above:
+
+1. **The night of the event**, once uploads have stopped. This is the backup that
+   matters; everything else is a formality.
+2. **After the host releases the gallery**, so the generated keepsake is captured too.
+3. Weekly thereafter, until the event is archived and torn down.
+
+Take the DB dump and the media tarball **back to back**, without uploads in flight
+between them. Upload rows reference files by path — a database from 22:00 and a media
+volume from 23:00 gives you rows pointing at files the dump doesn't know about, and
+rows whose files aren't in the tarball. Locking uploads from the host dashboard first
+(**Uploads sperren**) makes the pair genuinely consistent.
+
+---
+
+## Restore
+
+An untested backup is not a backup. Run this once against a scratch host **before**
+the event — it is roughly ten minutes, and it is the only way to find out that your
+tarball is empty or your dump is truncated while that is still a small problem.
+
+```bash
+# 0. Stop the app FIRST. Migrations run on boot and a live pool will fight the
+#    restore — a booting app against a half-restored schema can leave the migration
+#    table and the schema disagreeing, which is its own recovery problem.
+#    Leave `db` running: the dump is restored through it.
+docker compose stop app caddy
+
+# 1. Database. The dump carries its own DROPs (step 1 of Backup), so this replaces
+#    rather than collides. A dump taken WITHOUT --clean --if-exists will abort here
+#    on the first "already exists" — restore that one into a fresh empty database
+#    instead.
+gunzip -c ./backups/db_2026-07-29.sql.gz \
+  | docker compose exec -T db \
+      sh -c 'psql -U "$POSTGRES_USER" -d "$POSTGRES_DB" --set ON_ERROR_STOP=1'
+
+# 2. Media. NOTE the `--numeric-owner` and the chown: the app runs as a
+#    NON-ROOT user (uid 100, gid 101 — `addgroup -S app && adduser -S app`), and a
+#    restore that lands root-owned files makes every upload fail with EACCES deep in
+#    the write path, surfacing to the guest as a generic 500 with nothing in the UI
+#    to suggest permissions. The explicit chown is what guarantees it — BusyBox tar
+#    (which is what `alpine` ships) has no --same-owner, and restores ownership only
+#    because it runs as root here.
+docker run --rm \
+  -v eventsnap_media_data:/dst -v "$PWD/backups":/backup:ro \
+  alpine sh -c 'tar xzf /backup/media_2026-07-29.tar.gz -C /dst \
+                    --numeric-owner && chown -R 100:101 /dst'
+
+# 3. Exports. Same volume-name caveat, same ownership rules.
+docker run --rm \
+  -v eventsnap_exports_data:/dst -v "$PWD/backups":/backup:ro \
+  alpine sh -c 'tar xzf /backup/exports_2026-07-29.tar.gz -C /dst \
+                    --numeric-owner && chown -R 100:101 /dst'
+
+# 4. Back up. Migrations run, then export recovery re-arms any keepsake whose file
+#    didn't come back with the volume.
+docker compose up -d app caddy
+docker compose logs -f app        # watch for "migrations applied"
+
+# 5. Verify — all three, not just the first.
+curl -fsS https://DOMAIN/health && echo            # → ok
+#   … then sign in as host and confirm the feed renders images (proves the media
+#   volume restored AND is readable by uid 100), and that the keepsake downloads.
+```
+
+If the media volume restored but images 404 while the feed lists them, the paths are
+there and the bytes aren't — check `docker compose exec app ls -ln /media/originals`
+and confirm both the files and the `100:101` ownership.
+
+The restore is deliberately **not** automated. It is rare, destructive, and the one
+operation where a script that half-works is worse than a checklist someone reads.
 
 ---
 
@@ -254,7 +522,7 @@ Open:
 - [ ] SSE delta-fetch on foreground reconnect (scaffolded in [sse.ts](frontend/src/lib/sse.ts), not wired)
 - [ ] Live diashow / slideshow mode — see [docs/CONCEPT_DIASHOW.md](docs/CONCEPT_DIASHOW.md)
 - [ ] Individual file download button per post
-- [ ] Low-disk alert (< 10 GB free)
+- [x] Low-disk alert — host dashboard warns below 10 GB free, or whenever the keepsake would not fit
 - [ ] Event banner / cover image
 - [ ] Chunked resumable upload for files > 100 MB
 - [ ] Shared Tailwind config between main app and export-viewer

backend/.dockerignore

@@ -0,0 +1,12 @@
+# Build context exclusions.
+#
+# `target/` does not exist on a clean checkout, which is why builds have worked without this
+# file — but the moment anyone runs `cargo build` locally it becomes multi-GB, and every
+# `docker build` would ship all of it to the daemon for nothing (the Dockerfile only COPYs
+# Cargo.toml, Cargo.lock, src, static and migrations). On an EMULATED amd64 builder that
+# transfer is the slowest part of the build.
+target/
+
+# Never let a real .env reach an image layer.
+.env
+.env.*

backend/Cargo.lock

@@ -65,56 +65,6 @@ dependencies = [
  "libc",
 ]
 
-[[package]]
-name = "anstream"
-version = "1.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "824a212faf96e9acacdbd09febd34438f8f711fb84e09a8916013cd7815ca28d"
-dependencies = [
- "anstyle",
- "anstyle-parse",
- "anstyle-query",
- "anstyle-wincon",
- "colorchoice",
- "is_terminal_polyfill",
- "utf8parse",
-]
-
-[[package]]
-name = "anstyle"
-version = "1.0.14"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "940b3a0ca603d1eade50a4846a2afffd5ef57a9feac2c0e2ec2e14f9ead76000"
-
-[[package]]
-name = "anstyle-parse"
-version = "1.0.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "52ce7f38b242319f7cabaa6813055467063ecdc9d355bbb4ce0c68908cd8130e"
-dependencies = [
- "utf8parse",
-]
-
-[[package]]
-name = "anstyle-query"
-version = "1.1.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "40c48f72fd53cd289104fc64099abca73db4166ad86ea0b4341abe65af83dadc"
-dependencies = [
- "windows-sys 0.61.2",
-]
-
-[[package]]
-name = "anstyle-wincon"
-version = "3.0.11"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "291e6a250ff86cd4a820112fb8898808a366d8f9f58ce16d1f538353ad55747d"
-dependencies = [
- "anstyle",
- "once_cell_polyfill",
- "windows-sys 0.61.2",
-]
-
 [[package]]
 name = "anyhow"
 version = "1.0.102"
@@ -554,46 +504,12 @@ dependencies = [
  "inout",
 ]
 
-[[package]]
-name = "clap"
-version = "4.6.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b193af5b67834b676abd72466a96c1024e6a6ad978a1f484bd90b85c94041351"
-dependencies = [
- "clap_builder",
-]
-
-[[package]]
-name = "clap_builder"
-version = "4.6.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "714a53001bf66416adb0e2ef5ac857140e7dc3a0c48fb28b2f10762fc4b5069f"
-dependencies = [
- "anstream",
- "anstyle",
- "clap_lex",
- "strsim",
- "terminal_size",
-]
-
-[[package]]
-name = "clap_lex"
-version = "1.1.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "c8d4a3bb8b1e0c1050499d1815f5ab16d04f0959b233085fb31653fbfc9d98f9"
-
 [[package]]
 name = "color_quant"
 version = "1.1.0"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "3d7b894f5411737b7867f4827955924d7c254fc9f4d91a6aad6b097804b1018b"
 
-[[package]]
-name = "colorchoice"
-version = "1.0.5"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "1d07550c9036bf2ae0c684c4297d503f838287c83c53686d05370d0e139ae570"
-
 [[package]]
 name = "compression-codecs"
 version = "0.4.37"
@@ -677,15 +593,6 @@ dependencies = [
  "cfg-if",
 ]
 
-[[package]]
-name = "crossbeam-channel"
-version = "0.5.15"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "82b8f8f868b36967f9606790d1903570de9ceaf870a7bf9fbbd3016d636a2cb2"
-dependencies = [
- "crossbeam-utils",
-]
-
 [[package]]
 name = "crossbeam-deque"
 version = "0.8.6"
@@ -816,27 +723,6 @@ dependencies = [
  "cfg-if",
 ]
 
-[[package]]
-name = "env_filter"
-version = "1.0.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "32e90c2accc4b07a8456ea0debdc2e7587bdd890680d71173a15d4ae604f6eef"
-dependencies = [
- "log",
-]
-
-[[package]]
-name = "env_logger"
-version = "0.11.10"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0621c04f2196ac3f488dd583365b9c09be011a4ab8b9f37248ffcc8f6198b56a"
-dependencies = [
- "anstream",
- "anstyle",
- "env_filter",
- "log",
-]
-
 [[package]]
 name = "equator"
 version = "0.4.2"
@@ -1229,12 +1115,6 @@ dependencies = [
  "weezl",
 ]
 
-[[package]]
-name = "glob"
-version = "0.3.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "0cc23270f6e1808e30a928bdc84dea0b9b4136a8bc82338574f23baf47bbd280"
-
 [[package]]
 name = "governor"
 version = "0.6.3"
@@ -1622,7 +1502,6 @@ checksum = "7714e70437a7dc3ac8eb7e6f8df75fd8eb422675fc7678aff7364301092b1017"
 dependencies = [
  "equivalent",
  "hashbrown 0.16.1",
- "rayon",
  "serde",
  "serde_core",
 ]
@@ -1656,12 +1535,6 @@ dependencies = [
  "syn",
 ]
 
-[[package]]
-name = "is_terminal_polyfill"
-version = "1.70.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a6cb138bb79a146c1bd460005623e142ef0181e3d0219cb493e02f7d08a35695"
-
 [[package]]
 name = "itertools"
 version = "0.14.0"
@@ -1795,12 +1668,6 @@ dependencies = [
  "vcpkg",
 ]
 
-[[package]]
-name = "linux-raw-sys"
-version = "0.12.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "32a66949e030da00e8c7d4434b251670a91556f4144941d37452769c25d58a53"
-
 [[package]]
 name = "litemap"
 version = "0.8.1"
@@ -2089,12 +1956,6 @@ version = "1.21.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "9f7c3e4beb33f85d45ae3e3a1792185706c8e16d043238c593331cc7cd313b50"
 
-[[package]]
-name = "once_cell_polyfill"
-version = "1.70.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "384b8ab6d37215f3c5301a95a4accb5d64aa607f1fcb26a11b5303878451b4fe"
-
 [[package]]
 name = "oxipng"
 version = "9.1.5"
@@ -2102,18 +1963,12 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "26c613f0f566526a647c7473f6a8556dbce22c91b13485ee4b4ec7ab648e4973"
 dependencies = [
  "bitvec",
- "clap",
- "crossbeam-channel",
- "env_logger",
  "filetime",
- "glob",
  "indexmap",
  "libdeflater",
  "log",
- "rayon",
  "rgb",
  "rustc-hash",
- "zopfli",
 ]
 
 [[package]]
@@ -2607,19 +2462,6 @@ version = "2.1.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "94300abf3f1ae2e2b8ffb7b58043de3d399c73fa6f4b73826402a5c457614dbe"
 
-[[package]]
-name = "rustix"
-version = "1.1.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "b6fe4565b9518b83ef4f91bb47ce29620ca828bd32cb7e408f0062e9930ba190"
-dependencies = [
- "bitflags",
- "errno",
- "libc",
- "linux-raw-sys",
- "windows-sys 0.61.2",
-]
-
 [[package]]
 name = "rustversion"
 version = "1.0.22"
@@ -3060,12 +2902,6 @@ dependencies = [
  "unicode-properties",
 ]
 
-[[package]]
-name = "strsim"
-version = "0.11.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "7da8b5736845d9f2fcb837ea5d9e2628564b3b043a70948a3f0b778838c5fb4f"
-
 [[package]]
 name = "subtle"
 version = "2.6.1"
@@ -3120,16 +2956,6 @@ version = "1.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "55937e1799185b12863d447f42597ed69d9928686b8d88a1df17376a097d8369"
 
-[[package]]
-name = "terminal_size"
-version = "0.4.4"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "230a1b821ccbd75b185820a1f1ff7b14d21da1e442e22c0863ea5f08771a8874"
-dependencies = [
- "rustix",
- "windows-sys 0.61.2",
-]
-
 [[package]]
 name = "thiserror"
 version = "1.0.69"
@@ -3505,12 +3331,6 @@ version = "1.0.4"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b6c140620e7ffbb22c2dee59cafe6084a59b5ffc27a8859a5f0d494b5d52b6be"
 
-[[package]]
-name = "utf8parse"
-version = "0.2.2"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "06abde3611657adf66d383f00b093d7faecc7fa57071cce2578660c9f1010821"
-
 [[package]]
 name = "uuid"
 version = "1.23.0"
@@ -4187,18 +4007,6 @@ version = "1.0.21"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "b8848ee67ecc8aedbaf3e4122217aff892639231befc6a1b58d29fff4c2cabaa"
 
-[[package]]
-name = "zopfli"
-version = "0.8.3"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f05cd8797d63865425ff89b5c4a48804f35ba0ce8d125800027ad6017d2b5249"
-dependencies = [
- "bumpalo",
- "crc32fast",
- "log",
- "simd-adler32",
-]
-
 [[package]]
 name = "zstd"
 version = "0.13.3"

backend/Cargo.toml

@@ -27,7 +27,17 @@ tracing-subscriber = { version = "0.3", features = ["env-filter"] }
 dotenvy = "0.15"
 sysinfo = "0.32"
 image = "0.25"
-oxipng = "9"
+# default-features = false drops "parallel", which is what actually bounds oxipng's memory:
+# with rayon it evaluates row filters concurrently, each trial holding its own full-size
+# buffer, and there is no Options knob to cap that. Without the feature, lib.rs swaps in a
+# sequential shim (oxipng's own supported path) so peak scales with ONE trial, not N.
+# PNG optimisation gets slower; it is a background, best-effort, lossless size saving.
+#
+# "filetime" must be KEPT: without it OutFile::Path { preserve_attrs: true } silently no-ops.
+# Dropping "binary" also removes clap/glob/env_logger — a CLI's dependencies that were being
+# compiled into a server image — and "zopfli", which preset 2 does not use (it selects
+# Deflaters::Libdeflater, which is not feature-gated).
+oxipng = { version = "9", default-features = false, features = ["filetime"] }
 async_zip = { version = "0.0.17", features = ["tokio", "deflate"] }
 include_dir = "0.7"
 infer = "0.15"

backend/migrations/015_raise_upload_rate.down.sql

@@ -0,0 +1,3 @@
+-- Revert the default upload rate to 10/hour for installs still on the raised
+-- default (preserves any explicit admin override at another value).
+UPDATE config SET value = '10' WHERE key = 'upload_rate_per_hour' AND value = '100';

backend/migrations/015_raise_upload_rate.up.sql

@@ -0,0 +1,10 @@
+-- Raise the default per-guest upload rate from 10/hour to 100/hour.
+--
+-- Rationale: guests routinely upload a burst of 10-20 photos at once (phone
+-- multi-select). At the old default of 10/hour a real guest's first burst was
+-- throttled — surfaced by the 2026-07-18 load test. 100/hour comfortably covers
+-- several bursts across an event while still bounding abuse.
+--
+-- Only bump installs still on the old default; an admin who deliberately set a
+-- different value keeps it (migration 005 seeded 10; this UPDATE is scoped to '10').
+UPDATE config SET value = '100' WHERE key = 'upload_rate_per_hour' AND value = '10';

backend/migrations/016_display_derivative.down.sql

@@ -0,0 +1,28 @@
+-- Drop the view (frees the column dependency), remove the column, then restore the
+-- pre-016 view definition (matches migration 011).
+DROP VIEW IF EXISTS v_feed;
+ALTER TABLE upload DROP COLUMN display_path;
+
+CREATE VIEW v_feed AS
+SELECT
+    u.id,
+    u.event_id,
+    u.user_id,
+    usr.display_name             AS uploader_name,
+    usr.is_banned,
+    usr.uploads_hidden,
+    u.preview_path,
+    u.thumbnail_path,
+    u.mime_type,
+    u.caption,
+    u.created_at,
+    COUNT(DISTINCT l.user_id)    AS like_count,
+    COUNT(DISTINCT c.id)         AS comment_count
+FROM upload u
+JOIN "user" usr ON u.user_id = usr.id
+LEFT JOIN "like" l  ON l.upload_id = u.id
+LEFT JOIN comment c ON c.upload_id = u.id AND c.deleted_at IS NULL
+WHERE u.deleted_at IS NULL
+  AND usr.uploads_hidden = FALSE
+  AND usr.is_banned = FALSE
+GROUP BY u.id, usr.display_name, usr.is_banned, usr.uploads_hidden;

backend/migrations/016_display_derivative.up.sql

@@ -0,0 +1,34 @@
+-- Display derivative: a big-screen-quality image (~2048px long edge) for the diashow.
+-- The 800px `preview_path` is sized for phone feeds (data saver); upscaled on a projector
+-- it looks soft. The diashow uses `display_path` instead — bounded in size (safe to decode
+-- on weak kiosk hardware) yet sharp on 1080p/4K. NULL until the compression worker (or the
+-- one-time backfill) generates it; consumers fall back to the original when absent.
+ALTER TABLE upload ADD COLUMN display_path TEXT;
+
+-- Recreate (not CREATE OR REPLACE, which only allows appending columns at the end) so the
+-- new column can sit alongside preview_path/thumbnail_path.
+DROP VIEW IF EXISTS v_feed;
+CREATE VIEW v_feed AS
+SELECT
+    u.id,
+    u.event_id,
+    u.user_id,
+    usr.display_name             AS uploader_name,
+    usr.is_banned,
+    usr.uploads_hidden,
+    u.preview_path,
+    u.thumbnail_path,
+    u.display_path,
+    u.mime_type,
+    u.caption,
+    u.created_at,
+    COUNT(DISTINCT l.user_id)    AS like_count,
+    COUNT(DISTINCT c.id)         AS comment_count
+FROM upload u
+JOIN "user" usr ON u.user_id = usr.id
+LEFT JOIN "like" l  ON l.upload_id = u.id
+LEFT JOIN comment c ON c.upload_id = u.id AND c.deleted_at IS NULL
+WHERE u.deleted_at IS NULL
+  AND usr.uploads_hidden = FALSE
+  AND usr.is_banned = FALSE
+GROUP BY u.id, usr.display_name, usr.is_banned, usr.uploads_hidden;

backend/migrations/017_join_ip_rate.down.sql

@@ -0,0 +1 @@
+DELETE FROM config WHERE key IN ('join_ip_rate_per_min', 'admin_login_rate_enabled');

backend/migrations/017_join_ip_rate.up.sql

@@ -0,0 +1,18 @@
+-- Per-IP flood ceiling for /join, and the `admin_login_rate_enabled` toggle that
+-- every prior migration forgot to seed.
+--
+-- Rationale: /join was throttled at 5 requests per 60s keyed on the client IP. At a
+-- venue every guest is behind one NAT, so the whole party shared a single bucket —
+-- 12 guests scanning the QR code within a few seconds meant 5 got in and 7 were
+-- turned away. The handler now keys the real anti-spam bucket per (ip, name), the
+-- same shape as `recover:{ip}:{name}`, and keeps only a loose per-IP ceiling to bound
+-- raw volume. 60/min comfortably covers a whole wedding arriving at once while still
+-- capping a flood from a single source.
+--
+-- `admin_login_rate_enabled` is read by auth::handlers::admin_login with a code
+-- default of `true`, but no migration ever inserted it, so it was invisible to the
+-- admin config UI and to the e2e reseed. Seed it explicitly.
+INSERT INTO config (key, value) VALUES
+    ('join_ip_rate_per_min',     '60'),
+    ('admin_login_rate_enabled', 'true')
+ON CONFLICT (key) DO NOTHING;

backend/migrations/018_derivatives_rev.down.sql

@@ -0,0 +1,2 @@
+DROP INDEX IF EXISTS idx_upload_derivatives_rev;
+ALTER TABLE upload DROP COLUMN IF EXISTS derivatives_rev;

backend/migrations/018_derivatives_rev.up.sql

@@ -0,0 +1,18 @@
+-- Track which revision of the derivative pipeline produced an upload's preview/display.
+--
+-- Rev 1 applies the EXIF orientation tag. Everything generated before it decoded the raw
+-- sensor pixels and re-encoded to JPEG (which writes no EXIF), so every portrait phone photo
+-- was stored sideways in the feed preview, the diashow display and the keepsake — while the
+-- untouched original still rendered upright.
+--
+-- Existing rows default to 0 so the startup backfill can find and re-generate them exactly
+-- once; bump the constant in services/compression.rs if the pipeline ever changes again.
+ALTER TABLE upload ADD COLUMN IF NOT EXISTS derivatives_rev SMALLINT NOT NULL DEFAULT 0;
+
+-- Only image derivatives are affected — video thumbnails are extracted by ffmpeg, which
+-- already honours the rotation matrix. Mark them current so the backfill skips them.
+UPDATE upload SET derivatives_rev = 1 WHERE mime_type NOT LIKE 'image/%';
+
+CREATE INDEX IF NOT EXISTS idx_upload_derivatives_rev
+    ON upload (derivatives_rev)
+    WHERE deleted_at IS NULL;

backend/migrations/019_recover_ip_rate.down.sql

@@ -0,0 +1 @@
+DELETE FROM config WHERE key = 'recover_ip_rate_per_min';

backend/migrations/019_recover_ip_rate.up.sql

@@ -0,0 +1,19 @@
+-- Per-IP flood ceiling for /recover, mirroring the one migration 017 added for /join.
+--
+-- Rationale: /recover is keyed `recover:{ip}:{name}` at 5 per 15 minutes. That is the
+-- right shape for its actual job — stopping someone who knows a display name (they are
+-- visible on the feed) from burning the victim's 3-strike PIN counter and locking them
+-- out repeatedly. But the name is attacker-chosen, so cycling names mints a fresh bucket
+-- every time and the per-IP cost is unbounded.
+--
+-- Behind that limiter sits a cost-12 bcrypt verify, including an UNCONDITIONAL throwaway
+-- verify for names that don't exist — deliberately, to close a timing oracle. So an
+-- unknown name is the cheapest possible way to make the server do ~200ms of hashing.
+-- Without a ceiling, one client can saturate the box's CPU with a name generator.
+--
+-- 30/min is far above any real recovery attempt (a guest tries their PIN a handful of
+-- times) while capping a name-cycling flood. The per-(ip, name) bucket is unchanged and
+-- remains the anti-guessing control.
+INSERT INTO config (key, value) VALUES
+    ('recover_ip_rate_per_min', '30')
+ON CONFLICT (key) DO NOTHING;

backend/migrations/020_social_rate.down.sql

@@ -0,0 +1 @@
+DELETE FROM config WHERE key IN ('social_rate_per_min', 'social_rate_enabled');

backend/migrations/020_social_rate.up.sql

@@ -0,0 +1,16 @@
+-- Per-user rate limit for social writes (likes, comments, comment deletions).
+--
+-- These were the only writes in the app with no limit at all. Every other mutating
+-- path -- upload, join, recover, export, admin login -- carries one; social.rs
+-- carried none, so the coverage was asymmetric rather than deliberately open.
+--
+-- Severity is genuinely low for an invited-guest event, and the amplification worry
+-- turned out to be contained: a like fans an SSE broadcast to ~100 clients, but the
+-- export regeneration it could otherwise trigger is debounced (REGEN_DEBOUNCE 20s)
+-- and superseded workers are inert. So this closes the gap for symmetry, not urgency,
+-- and the ceiling is set high enough that no real guest will ever meet it -- a
+-- double-tapping enthusiast at a wedding is not the thing being defended against.
+INSERT INTO config (key, value) VALUES
+    ('social_rate_per_min',  '120'),
+    ('social_rate_enabled',  'true')
+ON CONFLICT (key) DO NOTHING;

backend/migrations/021_hashtag_counts_respect_bans.down.sql

@@ -0,0 +1,13 @@
+-- Restore the pre-021 definition (no ban/hide filtering) exactly as 004 created it.
+DROP VIEW IF EXISTS v_hashtag_counts;
+
+CREATE VIEW v_hashtag_counts AS
+SELECT
+    h.event_id,
+    h.tag,
+    COUNT(uh.upload_id) AS upload_count
+FROM hashtag h
+JOIN upload_hashtag uh ON uh.hashtag_id = h.id
+JOIN upload u          ON u.id = uh.upload_id AND u.deleted_at IS NULL
+GROUP BY h.event_id, h.id, h.tag
+ORDER BY upload_count DESC;

backend/migrations/021_hashtag_counts_respect_bans.up.sql

@@ -0,0 +1,31 @@
+-- v_hashtag_counts: apply the same visibility rules as v_feed.
+--
+-- The chip row and the grid's tag picker are both fed by this view, but it has only ever
+-- filtered `u.deleted_at IS NULL`. Migration 011 added ban/hide filtering to the feed and
+-- never reached here, so the two disagreed about which uploads exist:
+--
+--   * A host bans a guest who posted 3 of the 12 `#tanz` photos. The chip keeps reading
+--     "#tanz 12"; tapping it returns 9. The count is presented as authoritative and is not.
+--   * A tag used ONLY by a banned or hidden guest stays in the chip row and in the tag
+--     picker as a selectable option that leads to an empty feed — a ghost filter that
+--     cannot be cleared because there is nothing wrong with it to see.
+--
+-- Bans are exactly the moment a host is watching these numbers to confirm the moderation
+-- took effect, so a stale count reads as "the ban didn't work".
+--
+-- Same predicate as v_feed (see 016_display_derivative.up.sql), joined through `user`.
+DROP VIEW IF EXISTS v_hashtag_counts;
+
+CREATE VIEW v_hashtag_counts AS
+SELECT
+    h.event_id,
+    h.tag,
+    COUNT(uh.upload_id) AS upload_count
+FROM hashtag h
+JOIN upload_hashtag uh ON uh.hashtag_id = h.id
+JOIN upload u          ON u.id = uh.upload_id AND u.deleted_at IS NULL
+JOIN "user" usr        ON usr.id = u.user_id
+WHERE usr.uploads_hidden = FALSE
+  AND usr.is_banned = FALSE
+GROUP BY h.event_id, h.id, h.tag
+ORDER BY upload_count DESC;

backend/migrations/022_client_upload_idempotency.down.sql

@@ -0,0 +1,2 @@
+DROP INDEX IF EXISTS upload_client_upload_id_key;
+ALTER TABLE upload DROP COLUMN IF EXISTS client_upload_id;

backend/migrations/022_client_upload_idempotency.up.sql

@@ -0,0 +1,25 @@
+-- Idempotency key for uploads, supplied by the client.
+--
+-- The failure this closes is the ordinary one on a phone, not an exotic race: the server
+-- receives the body, validates it, commits the row, and the response is lost on the way back
+-- because the guest walked out of range or the AP dropped the connection. The client sees a
+-- network error with the blob still in hand, marks the item retryable, and re-sends it — both
+-- when the guest taps "Erneut" and automatically when the queue requeues on reconnect. Every
+-- attempt minted a fresh `Uuid::new_v4()` server-side, so the same photo landed in the gallery
+-- two or three times and was charged against the guest's storage quota each time.
+--
+-- The client already has a stable per-queue-item UUID, so it costs nothing to send. NULL is
+-- allowed and unconstrained: uploads that predate this column, and any client that doesn't send
+-- one, keep working exactly as before.
+ALTER TABLE upload ADD COLUMN client_upload_id UUID;
+
+-- Partial rather than a plain UNIQUE. Postgres would tolerate the NULLs either way, but indexing
+-- only the rows that carry a key keeps it small and states the rule exactly: uniqueness applies
+-- where a key exists, and nowhere else.
+--
+-- Scoped globally rather than per user or per event. The key is a client-generated v4 UUID, so a
+-- collision between two different photos is not a real possibility, and a single-column index
+-- means the uniqueness check cannot be wrong about which event or user a retry belongs to.
+CREATE UNIQUE INDEX upload_client_upload_id_key
+    ON upload (client_upload_id)
+    WHERE client_upload_id IS NOT NULL;

backend/migrations/023_derivative_attempts.down.sql

@@ -0,0 +1,3 @@
+DROP INDEX IF EXISTS idx_upload_derivative_backfill;
+ALTER TABLE upload DROP COLUMN IF EXISTS derivative_last_error;
+ALTER TABLE upload DROP COLUMN IF EXISTS derivative_attempts;

backend/migrations/023_derivative_attempts.up.sql

@@ -0,0 +1,23 @@
+-- Bound how many times a permanently-failing upload can be re-processed.
+--
+-- Without this, one poisoned row is an outage. The upload row is committed BEFORE compression
+-- starts, `derivatives_rev` defaults to 0, and `set_derivatives_rev` only runs on success — so
+-- a row whose processing kills the container survives at rev 0, the unconditional startup
+-- backfill re-selects it on the next boot, and `restart: unless-stopped` turns that into an
+-- infinite kill loop. Every restart also drops every SSE stream and truncates every in-flight
+-- upload. That was reachable via a single large PNG (see services/compression.rs), but the
+-- shape is general: any input that can kill or hang the worker repeats forever.
+--
+-- The counter is incremented WRITE-AHEAD, before the work is attempted, because the failure
+-- mode being defended against is a SIGKILL — no error is returned, no handler runs, no Drop
+-- fires. A counter bumped in an error path increments zero times per crash and changes nothing.
+ALTER TABLE upload ADD COLUMN IF NOT EXISTS derivative_attempts SMALLINT NOT NULL DEFAULT 0;
+
+-- Last failure text, so a row that has given up can be diagnosed without reproducing it.
+-- Nothing reads this in code; it exists for the operator.
+ALTER TABLE upload ADD COLUMN IF NOT EXISTS derivative_last_error TEXT;
+
+-- Serves the backfill selection, which now filters on both columns.
+CREATE INDEX IF NOT EXISTS idx_upload_derivative_backfill
+    ON upload (derivatives_rev, derivative_attempts)
+    WHERE deleted_at IS NULL;

backend/migrations/024_feed_scalar_counts.down.sql

@@ -0,0 +1,26 @@
+-- Restore the 016 definition verbatim.
+DROP VIEW IF EXISTS v_feed;
+CREATE VIEW v_feed AS
+SELECT
+    u.id,
+    u.event_id,
+    u.user_id,
+    usr.display_name             AS uploader_name,
+    usr.is_banned,
+    usr.uploads_hidden,
+    u.preview_path,
+    u.thumbnail_path,
+    u.display_path,
+    u.mime_type,
+    u.caption,
+    u.created_at,
+    COUNT(DISTINCT l.user_id)    AS like_count,
+    COUNT(DISTINCT c.id)         AS comment_count
+FROM upload u
+JOIN "user" usr ON u.user_id = usr.id
+LEFT JOIN "like" l  ON l.upload_id = u.id
+LEFT JOIN comment c ON c.upload_id = u.id AND c.deleted_at IS NULL
+WHERE u.deleted_at IS NULL
+  AND usr.uploads_hidden = FALSE
+  AND usr.is_banned = FALSE
+GROUP BY u.id, usr.display_name, usr.is_banned, usr.uploads_hidden;

backend/migrations/024_feed_scalar_counts.up.sql

@@ -0,0 +1,55 @@
+-- Make a feed page cost a page, not the whole event.
+--
+-- The previous definition (016) computed like_count/comment_count with LEFT JOINs and a
+-- GROUP BY. Postgres CAN push the `event_id = $1` qual and the keyset predicate through the
+-- view — verified with EXPLAIN, it uses idx_upload_event_created_id — but it CANNOT push
+-- ORDER BY ... LIMIT across a GroupAggregate. So every feed request aggregated every upload in
+-- the event (times its likes and comments) and only then sorted and took 21 rows. The cost
+-- grew with the event, not with the page, and page 1 — the most expensive one — is exactly
+-- what refreshFeedInPlace refetches on every completed upload, from every open feed in the
+-- venue.
+--
+-- Correlated scalar subqueries move the counts ABOVE the Limit in the plan: they are evaluated
+-- once per returned row, so 21 index lookups instead of a full aggregation.
+--
+-- The rewrite is EXACTLY equivalent, not merely close:
+--   * "like" is keyed (upload_id, user_id), so COUNT(DISTINCT l.user_id) == count(*).
+--   * comment.id is the primary key, so COUNT(DISTINCT c.id) == count(*).
+--   * one row per upload either way — the GROUP BY was on u.id.
+-- Column names, order and types are unchanged (count(*) and COUNT(DISTINCT ...) are both
+-- bigint), so no Rust code changes.
+--
+-- No new index needed: idx_like_upload plus the (upload_id, user_id) PK serve the like
+-- subquery, and idx_comment_upload ... WHERE deleted_at IS NULL matches the comment
+-- subquery's predicate exactly.
+--
+-- One thing a future editor needs to know: the hashtag-filtered feed joins upload_hashtag
+-- against this view. That was safe before only because the GROUP BY collapsed the join
+-- fan-out; it is safe now because the view is one row per upload and upload_hashtag is keyed
+-- (upload_id, hashtag_id) with a single tag filtered. Adding a second tag filter would need
+-- fresh thought.
+
+-- Not CASCADE: if something ever comes to depend on this view, the migration should fail
+-- loudly rather than silently drop it.
+DROP VIEW IF EXISTS v_feed;
+CREATE VIEW v_feed AS
+SELECT
+    u.id,
+    u.event_id,
+    u.user_id,
+    usr.display_name AS uploader_name,
+    usr.is_banned,
+    usr.uploads_hidden,
+    u.preview_path,
+    u.thumbnail_path,
+    u.display_path,
+    u.mime_type,
+    u.caption,
+    u.created_at,
+    (SELECT count(*) FROM "like" l WHERE l.upload_id = u.id)                           AS like_count,
+    (SELECT count(*) FROM comment c WHERE c.upload_id = u.id AND c.deleted_at IS NULL) AS comment_count
+FROM upload u
+JOIN "user" usr ON u.user_id = usr.id
+WHERE u.deleted_at IS NULL
+  AND usr.uploads_hidden = FALSE
+  AND usr.is_banned = FALSE;

backend/migrations/025_reserved_names_and_pin_decay.down.sql

@@ -0,0 +1,11 @@
+-- NOTE: the reserved-name rename in the up migration is NOT reversible. The original names
+-- are not recorded anywhere, and reversing it would in any case re-create the state that
+-- bricked admin login. Rolling back the schema does not roll back that data change.
+DELETE FROM config WHERE key IN (
+    'recover_name_rate_per_15min',
+    'pin_reset_ip_rate_per_min',
+    'upload_edit_rate_per_min',
+    'upload_edit_rate_enabled'
+);
+
+ALTER TABLE "user" DROP COLUMN IF EXISTS last_failed_pin_at;

backend/migrations/025_reserved_names_and_pin_decay.up.sql

@@ -0,0 +1,70 @@
+-- Two independent auth defects that share a migration because they share a table.
+
+-- 1. RESERVED NAMES — free any guest squatting on a name the admin path used to depend on.
+--
+-- Migration 007 made display_name unique per event case-insensitively, and `join` had no
+-- reserved-name guard. So any guest could join as "admin"/"Admin"/"ADMIN" before the operator's
+-- first admin login; admin_login then looked its user up BY NAME, missed (wrong role), fell
+-- through to creating "Admin", violated that unique index, and returned a 500 — permanently,
+-- with no in-app recovery. Moderation, config and gallery release all gone, fixed only by SQL.
+--
+-- The real fix is in code (look the admin up by role, never by name — see auth/handlers.rs).
+-- This clears the state an already-deployed database may be carrying.
+--
+-- RENAMED, NEVER DELETED: the guest keeps their uploads, their PIN and their session. Only
+-- non-admin rows are touched — a real admin row named "Admin" is the expected state.
+-- Two guards that are not optional, because this statement runs INSIDE the migration
+-- transaction on boot and a failure here exits the process — `restart: unless-stopped` then
+-- turns it into a crash loop with no in-app recovery. That is a strictly worse version of the
+-- lockout this migration exists to clean up after.
+--
+--   * role = 'guest', not role <> 'admin'. The enum also has 'host' (001), and hosts are
+--     promoted from guests at runtime — so <> 'admin' renamed a legitimately promoted staff
+--     member whose name happens to be "Host".
+--   * NOT EXISTS. The target name is derived, not unique: `idx_user_event_name_ci` (007) is a
+--     UNIQUE index on (event_id, lower(display_name)), and nothing stopped a second guest from
+--     having already joined as exactly "Admin (a1b2c3d4)" — the old code had no reserved-name
+--     guard and the join response hands each guest their own id. Rare, but the cost of losing
+--     that bet is the whole event.
+--
+-- A row that collides is simply left alone: `create_admin_user` already handles a name clash by
+-- falling back to `Admin-<8hex>`, and `admin_login` no longer resolves by name at all, so this
+-- cleanup is convenience rather than load-bearing.
+UPDATE "user" u
+   SET display_name = u.display_name || ' (' || left(u.id::text, 8) || ')'
+ WHERE u.role = 'guest'
+   AND lower(u.display_name) IN ('admin', 'administrator', 'host', 'eventsnap')
+   AND NOT EXISTS (
+        SELECT 1 FROM "user" x
+         WHERE x.event_id = u.event_id
+           AND lower(x.display_name) = lower(u.display_name || ' (' || left(u.id::text, 8) || ')')
+   );
+
+-- 2. PIN LOCKOUT DECAY.
+--
+-- failed_pin_attempts only ever cleared on a successful recovery or after a lockout expired, so
+-- honest typos accumulated across days: a guest who fat-fingered their PIN twice last night
+-- arrives today already two-thirds of the way to being locked out. With the threshold now
+-- raised (see below) a decay window is what keeps that raise safe rather than merely lenient.
+ALTER TABLE "user" ADD COLUMN IF NOT EXISTS last_failed_pin_at TIMESTAMPTZ;
+
+-- Rate-limit knobs introduced with this release.
+--
+-- recover_name_rate_per_15min (4, was a hardcoded 5): the per-(IP, name) ceiling. It MUST stay
+-- below the account-lock threshold, which is the whole defect — at 5-per-IP against a 3-strike
+-- lock, three requests from one IP locked any guest whose name is visible on the feed, every 15
+-- minutes, forever. The lock threshold moves to 12 in code, so locking a victim now needs at
+-- least three distinct sources while an honest guest never comes close.
+--
+-- pin_reset_ip_rate_per_min (30): /recover/request was the one unauthenticated endpoint with no
+-- per-IP ceiling at all — /join got one in 017 and /recover in 019, and this third one was
+-- simply missed. Its per-name key is attacker-chosen, so cycling names minted a fresh bucket
+-- every time and the per-IP cost was unbounded.
+--
+-- upload_edit_rate_per_min (30): PATCH /upload/{id} had no rate limit of any kind.
+INSERT INTO config (key, value) VALUES
+    ('recover_name_rate_per_15min', '4'),
+    ('pin_reset_ip_rate_per_min',   '30'),
+    ('upload_edit_rate_per_min',    '30'),
+    ('upload_edit_rate_enabled',    'true')
+ON CONFLICT (key) DO NOTHING;

backend/src/auth/handlers.rs

@@ -1,11 +1,12 @@
 use std::time::Duration;
 
 use axum::Json;
-use axum::extract::State;
+use axum::extract::{ConnectInfo, State};
 use axum::http::{HeaderMap, StatusCode};
 use chrono::Utc;
 use rand::Rng;
 use serde::{Deserialize, Serialize};
+use std::net::SocketAddr;
 use uuid::Uuid;
 
 use crate::auth::jwt;
@@ -18,6 +19,58 @@ use crate::services::config;
 use crate::services::rate_limiter::client_ip;
 use crate::state::AppState;
 
+/// Names a guest may not take.
+///
+/// Defence in depth only. The real fix for the admin-lockout defect is that `admin_login` now
+/// resolves its user by ROLE rather than by name (see `User::find_admin_for_event`), which is
+/// why homoglyph and zero-width bypasses of this list are not a concern: the name is no longer
+/// load-bearing for anything. What this buys is that a guest cannot impersonate the host in the
+/// feed's byline, and that "Admin" stays available for the admin row.
+const RESERVED_DISPLAY_NAMES: &[&str] = &["admin", "administrator", "host", "eventsnap"];
+
+fn is_reserved_display_name(name: &str) -> bool {
+    let name = name.trim().to_lowercase();
+    RESERVED_DISPLAY_NAMES.contains(&name.as_str())
+}
+
+/// Trim and bounds-check a display name.
+///
+/// Shared by `join`, `recover` and `request_pin_reset` so the length check happens BEFORE the
+/// name is used to build a rate-limiter key. It was inline in `join` only, so on the other two
+/// endpoints `format!("...:{ip}:{name_key}")` allocated from an unbounded, attacker-chosen
+/// string and stored it in a HashMap pruned once an hour with a 24 h ceiling — turning the
+/// limiter itself into the memory-exhaustion primitive it exists to prevent.
+fn validate_display_name(raw: &str) -> Result<&str, AppError> {
+    let name = raw.trim();
+    let chars = name.chars().count();
+    if chars == 0 || chars > 50 {
+        return Err(AppError::BadRequest(
+            "Name muss zwischen 1 und 50 Zeichen lang sein.".into(),
+        ));
+    }
+    // No control characters. NUL is the hard requirement — Postgres rejects 0x00 in TEXT with a
+    // 500, so catching it here turns an internal error into a clean 400 — but the rest matter
+    // too, and for reasons beyond tidiness:
+    //
+    //  * Newlines make the name a LOG INJECTION vector. Several 4xx messages interpolate it
+    //    ("Der Name \"X\" ist bereits vergeben.") and those are logged; a name carrying a
+    //    newline plus a plausible timestamp prefix lets two unauthenticated requests forge
+    //    entries in the only forensic record an unattended event has. `error.rs` escapes on the
+    //    way out as well — this is the other half, and the half that keeps the forged text out
+    //    of the database and out of the feed byline in the first place.
+    //  * A bare CR or a bidi override renders as a name that is not what was typed, in the feed,
+    //    the host dashboard's moderation list and the keepsake.
+    //
+    // Deliberately NOT a whitelist: guests have accents, emoji and non-Latin scripts in their
+    // names, and rejecting those would be worse than the problem.
+    if name.chars().any(|c| c.is_control()) {
+        return Err(AppError::BadRequest(
+            "Name enthält ungültige Zeichen.".into(),
+        ));
+    }
+    Ok(name)
+}
+
 #[derive(Deserialize)]
 pub struct JoinRequest {
     pub display_name: String,
@@ -33,37 +86,58 @@ pub struct JoinResponse {
 
 pub async fn join(
     State(state): State<AppState>,
+    ConnectInfo(peer): ConnectInfo<SocketAddr>,
     headers: HeaderMap,
     Json(body): Json<JoinRequest>,
 ) -> Result<(StatusCode, Json<JoinResponse>), AppError> {
-    let ip = client_ip(&headers, "unknown");
+    let ip = client_ip(&headers, &peer.ip().to_string());
     let rate_limits_on = config::get_bool(&state.config_cache, "rate_limits_enabled", true).await;
     let join_rate_on = config::get_bool(&state.config_cache, "join_rate_enabled", true).await;
-    if rate_limits_on
-        && join_rate_on
-        && !state
-            .rate_limiter
-            .check(format!("join:{ip}"), 5, Duration::from_secs(60))
-    {
-        return Err(AppError::TooManyRequests(
-            "Zu viele Anfragen. Bitte warte kurz und versuche es erneut.".into(),
-            None,
-        ));
+
+    // Coarse per-IP flood ceiling. `/join` is pre-auth so there is no user to key on, and
+    // at a venue EVERY guest arrives from one public IP — a tight per-IP bucket meant the
+    // 6th person through the door was turned away by the 5 ahead of them. So the per-IP
+    // limit here only bounds raw volume; the real anti-spam bucket is per-name below.
+    // Cheap enough to run before validation, which keeps a flood of malformed bodies from
+    // being free.
+    if rate_limits_on && join_rate_on {
+        let ip_ceiling = config::get_usize(&state.config_cache, "join_ip_rate_per_min", 60).await;
+        if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
+            format!("join_ip:{ip}"),
+            ip_ceiling,
+            Duration::from_secs(60),
+        ) {
+            return Err(AppError::TooManyRequests(
+                "Zu viele Anfragen. Bitte warte kurz und versuche es erneut.".into(),
+                Some(retry_after_secs),
+            ));
+        }
     }
 
-    let display_name = body.display_name.trim();
-    let name_chars = display_name.chars().count();
-    if name_chars == 0 || name_chars > 50 {
-        return Err(AppError::BadRequest(
-            "Name muss zwischen 1 und 50 Zeichen lang sein.".into(),
-        ));
+    let display_name = validate_display_name(&body.display_name)?;
+    if is_reserved_display_name(display_name) {
+        // 409, matching the name-taken response below, so the frontend's existing handling
+        // works unchanged. See RESERVED_DISPLAY_NAMES for why this exists.
+        return Err(AppError::Conflict(format!(
+            "Der Name \"{display_name}\" ist reserviert. Bitte wähle einen anderen."
+        )));
     }
-    // Postgres rejects 0x00 in TEXT columns with a 500. Catch it here so callers
-    // see a clean 400 instead of an internal error.
-    if display_name.contains('\0') {
-        return Err(AppError::BadRequest(
-            "Name enthält ungültige Zeichen.".into(),
-        ));
+
+    // Per-guest bucket, keyed like the `recover:{ip}:{name}` limiter below. This carries
+    // the original 5/60s anti-spam intent, but one guest retrying can no longer consume
+    // the allowance of everyone else sharing the venue's NAT.
+    if rate_limits_on && join_rate_on {
+        let name_key = display_name.to_lowercase();
+        if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
+            format!("join:{ip}:{name_key}"),
+            5,
+            Duration::from_secs(60),
+        ) {
+            return Err(AppError::TooManyRequests(
+                "Zu viele Anfragen. Bitte warte kurz und versuche es erneut.".into(),
+                Some(retry_after_secs),
+            ));
+        }
     }
 
     let event = Event::find_or_create(
@@ -83,7 +157,7 @@ pub async fn join(
 
     // Generate a 4-digit PIN
     let pin: String = format!("{:04}", rand::rng().random_range(0..10000u32));
-    let pin_hash = bcrypt::hash(&pin, 12).map_err(|e| AppError::Internal(anyhow::anyhow!(e)))?;
+    let pin_hash = hash_password(pin.clone(), 12).await?;
 
     // The pre-check above is racy: two simultaneous joins with the same name can both
     // pass it, and the DB's unique index then rejects the loser. Map that unique
@@ -123,6 +197,38 @@ pub async fn join(
     ))
 }
 
+/// Default for `recover_name_rate_per_15min` — wrong PINs allowed per (IP, name) per 15 min.
+/// Mirrors migration 023; kept here so the invariant below can be asserted in a test.
+const RECOVER_NAME_CEILING_DEFAULT: usize = 4;
+
+/// Hard ceiling on the CONFIGURED per-(IP, name) limit, whatever an operator sets.
+///
+/// The ordering `3 x ceiling <= PIN_LOCK_THRESHOLD` is the entire control that stops one source
+/// from locking a victim out: display names are public on the feed, so if a single IP can spend
+/// the whole lock threshold it can lock any guest it likes, repeatedly. That ordering was asserted
+/// in a comment and in a test — but the test pinned the DEFAULT constant, while the handler reads
+/// the config value, and `patch_config` accepted anything from 1 to 100_000. So an operator
+/// raising this key restored the exact DoS the tier ordering exists to prevent, silently.
+pub const RECOVER_NAME_CEILING_MAX: usize = (PIN_LOCK_THRESHOLD as usize) / 3;
+
+/// Wrong PINs, from ALL sources, before the ACCOUNT itself is locked for 15 minutes.
+///
+/// This was 3, which sat BELOW the per-(IP, name) ceiling of 5 — and that ordering, not the
+/// number, was the defect. Display names are public on the feed, so three requests from a single
+/// IP locked any guest out of their own account, repeatable every 15 minutes, indefinitely. The
+/// tier meant to protect a guest was the easiest way to attack them.
+///
+/// Raised deliberately far above the per-IP tier so the two do different jobs. The per-(IP, name)
+/// bucket is what stops a guesser, and it costs the ATTACKER. This tier is the last line against
+/// a DISTRIBUTED guesser, and it is the only one an attacker can turn on a victim — so reaching
+/// it must require at least three distinct sources inside the decay window.
+///
+/// Brute-force cost is unchanged: 12 attempts per 15 minutes is 48/hour against one account, so
+/// 10 000 four-digit PINs still take ~208 hours no matter how many IPs are used. An honest guest
+/// fat-fingering a 4-digit PIN never comes close, and `increment_failed_pin` now decays the
+/// streak after 15 minutes so yesterday's typos don't count toward today's.
+const PIN_LOCK_THRESHOLD: i16 = 12;
+
 #[derive(Deserialize)]
 pub struct RecoverRequest {
     pub display_name: String,
@@ -147,31 +253,114 @@ fn dummy_pin_hash() -> &'static str {
     })
 }
 
+/// Run a bcrypt verify on the blocking pool.
+///
+/// bcrypt at cost 12 is ~200ms of deliberate CPU. Called inline on an async task it pins a
+/// tokio WORKER thread for that whole time, and the runtime only has one per core — so a
+/// flood of `/recover` or `/admin/login` attempts stalls every other request on the box,
+/// including the feed. Offloading moves that cost to the blocking pool, which is sized for
+/// exactly this and whose saturation degrades logins rather than the whole app.
+/// Process-wide ceiling on CONCURRENT bcrypt work.
+///
+/// bcrypt is deliberately expensive — ~250 ms of a core at cost 12, and this deployment's own
+/// runbook generates the admin hash at a higher cost than that. Every call is correctly on
+/// `spawn_blocking`, but tokio's blocking pool defaults to 512 threads, so "off the async
+/// runtime" is not the same as "bounded": enough concurrent hashes will preempt both async
+/// worker threads a 2 vCPU box gets, and uploads, feed and SSE stall behind them.
+///
+/// Three unauthenticated endpoints reach bcrypt — `/join` (hash), `/recover` (verify, including
+/// a deliberate throwaway verify for unknown names) and `/admin/login` (verify) — each with only
+/// a per-IP bucket in front, and at a venue every guest shares one public IP. A per-IP limit
+/// therefore bounds nothing globally.
+///
+/// `cores - 1` leaves a core for actually serving requests. Excess callers WAIT on the permit
+/// rather than burning CPU, so a flood degrades to latency instead of an outage.
+static BCRYPT_PERMITS: std::sync::LazyLock<tokio::sync::Semaphore> =
+    std::sync::LazyLock::new(|| {
+        let cores = std::thread::available_parallelism()
+            .map(|n| n.get())
+            .unwrap_or(2);
+        tokio::sync::Semaphore::new(cores.saturating_sub(1).max(1))
+    });
+
+async fn verify_password(candidate: String, hash: String) -> bool {
+    // `acquire()` only fails if the semaphore is closed, which never happens here.
+    let _permit = BCRYPT_PERMITS.acquire().await;
+    tokio::task::spawn_blocking(move || bcrypt::verify(&candidate, &hash).unwrap_or(false))
+        .await
+        .unwrap_or(false)
+}
+
+/// Hash a secret on the blocking pool. Same reasoning as [`verify_password`] — and this one
+/// runs on the busiest auth path there is, since every guest who joins gets a PIN hashed.
+pub async fn hash_password(secret: String, cost: u32) -> Result<String, AppError> {
+    // Same global ceiling as `verify_password` — `/join` hashes a PIN for every guest, and 100
+    // guests scanning the QR at once is the arrival burst this box has to survive.
+    let _permit = BCRYPT_PERMITS.acquire().await;
+    tokio::task::spawn_blocking(move || bcrypt::hash(&secret, cost))
+        .await
+        .map_err(|e| AppError::Internal(anyhow::anyhow!(e)))?
+        .map_err(|e| AppError::Internal(anyhow::anyhow!(e)))
+}
+
 pub async fn recover(
     State(state): State<AppState>,
+    ConnectInfo(peer): ConnectInfo<SocketAddr>,
     headers: HeaderMap,
     Json(body): Json<RecoverRequest>,
 ) -> Result<Json<RecoverResponse>, AppError> {
-    let display_name = body.display_name.trim();
+    // Validated BEFORE it is used as a rate-limiter key — see `validate_display_name`. The
+    // per-IP ceiling below is keyed only on the IP, so it is safe to run either side of this;
+    // the per-NAME bucket is not.
+    let display_name = validate_display_name(&body.display_name)?;
 
-    // Per-IP+name throttle BEFORE the per-user 3-strike counter. Without this
-    // an attacker who knows a display name (they're visible on the feed) can
-    // burn through 3 wrong PINs and lock the victim for 15 minutes — repeated
-    // every 15 minutes, indefinitely. 5 attempts per 15 minutes per (IP, name)
-    // softens that into a real cost.
-    let ip = client_ip(&headers, "unknown");
+    // Per-IP+name throttle BEFORE the per-user lockout counter. Without this an attacker who
+    // knows a display name (they're visible on the feed) can burn through the victim's wrong-PIN
+    // budget and lock them out, repeatedly. The ceiling here MUST stay below
+    // PIN_LOCK_THRESHOLD — see the constant for why that ordering is the whole control.
+    let ip = client_ip(&headers, &peer.ip().to_string());
     let rate_limits_on = config::get_bool(&state.config_cache, "rate_limits_enabled", true).await;
     let recover_rate_on = config::get_bool(&state.config_cache, "recover_rate_enabled", true).await;
     if rate_limits_on && recover_rate_on {
+        // Coarse per-IP ceiling FIRST. The per-(ip, name) bucket below is the anti-guessing
+        // control, but the name is attacker-chosen, so cycling names mints a fresh bucket
+        // every time and leaves the per-IP cost unbounded. That matters more here than
+        // anywhere else: every call runs a cost-12 bcrypt verify, including an
+        // unconditional throwaway one for names that don't exist (see below), so an unknown
+        // name is the CHEAPEST way to make the server do ~200ms of hashing. Checked before
+        // the per-name bucket so a name generator can't walk past it.
+        let ip_ceiling =
+            config::get_usize(&state.config_cache, "recover_ip_rate_per_min", 30).await;
+        if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
+            format!("recover_ip:{ip}"),
+            ip_ceiling,
+            Duration::from_secs(60),
+        ) {
+            return Err(AppError::TooManyRequests(
+                "Zu viele Versuche. Bitte warte kurz und versuche es erneut.".into(),
+                Some(retry_after_secs),
+            ));
+        }
+
+        let name_ceiling = config::get_usize(
+            &state.config_cache,
+            "recover_name_rate_per_15min",
+            RECOVER_NAME_CEILING_DEFAULT,
+        )
+        .await
+        // CLAMPED, not merely defaulted — see RECOVER_NAME_CEILING_MAX. The config value is
+        // operator-settable and the invariant it has to respect is not expressible in
+        // `patch_config`'s numeric range, so it is enforced at the point of use.
+        .min(RECOVER_NAME_CEILING_MAX);
         let name_key = display_name.to_lowercase();
-        if !state.rate_limiter.check(
+        if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
             format!("recover:{ip}:{name_key}"),
-            5,
+            name_ceiling,
             Duration::from_secs(15 * 60),
         ) {
             return Err(AppError::TooManyRequests(
                 "Zu viele Versuche. Bitte warte kurz und versuche es erneut.".into(),
-                None,
+                Some(retry_after_secs),
             ));
         }
     }
@@ -188,7 +377,7 @@ pub async fn recover(
         // PIN — so "no such name" and "wrong PIN" are indistinguishable by response or
         // timing. Display names are already public on the feed, but this still closes
         // the /recover enumeration + timing oracle.
-        let _ = bcrypt::verify(&body.pin, dummy_pin_hash());
+        let _ = verify_password(body.pin.clone(), dummy_pin_hash().to_string()).await;
         return Err(AppError::Unauthorized("PIN ist falsch.".into()));
     }
 
@@ -200,16 +389,19 @@ pub async fn recover(
         // is effectively permanently fragile.
         if let Some(locked_until) = user.pin_locked_until {
             if Utc::now() < locked_until {
+                // The exact deadline is known, so surface it as Retry-After instead of
+                // making the client guess at the "15 Minuten" in the copy.
+                let retry_after_secs = (locked_until - Utc::now()).num_seconds().max(1) as u64;
                 return Err(AppError::TooManyRequests(
                     "Zu viele Versuche. Bitte warte 15 Minuten.".into(),
-                    None,
+                    Some(retry_after_secs),
                 ));
             }
             // Lockout window expired — wipe the counter and the timestamp.
             User::reset_pin_attempts(&state.pool, user.id).await?;
         }
 
-        let pin_matches = bcrypt::verify(&body.pin, &user.recovery_pin_hash).unwrap_or(false);
+        let pin_matches = verify_password(body.pin.clone(), user.recovery_pin_hash.clone()).await;
 
         if pin_matches {
             // Reset failed attempts on success
@@ -243,7 +435,7 @@ pub async fn recover(
             attempts,
             "recover: wrong PIN"
         );
-        if attempts >= 3 {
+        if attempts >= PIN_LOCK_THRESHOLD {
             let lockout = Utc::now() + chrono::Duration::minutes(15);
             User::lock_pin(&state.pool, user.id, lockout).await?;
             tracing::warn!(
@@ -272,8 +464,21 @@ pub struct AdminLoginResponse {
     pub display_name: String,
 }
 
+/// Requests per minute per IP that may reach `verify_password` at all.
+///
+/// Not a security control — the failure bucket below is. It bounds how deep a queue can form on
+/// `BCRYPT_PERMITS`, which is what actually caps the CPU cost.
+///
+/// Still far above anything a person typing a password produces, but note the honest limitation:
+/// unlike the failure bucket, this ceiling CAN refuse a correct password, and on venue NAT every
+/// guest shares the operator's IP. It is a smaller number than it first was for exactly that
+/// reason — the earlier 120 was chosen when this was the only bound on bcrypt, which made it both
+/// too weak to cap CPU and too coarse to be safe for the operator.
+const ADMIN_LOGIN_CPU_CEILING: usize = 30;
+
 pub async fn admin_login(
     State(state): State<AppState>,
+    ConnectInfo(peer): ConnectInfo<SocketAddr>,
     headers: HeaderMap,
     Json(body): Json<AdminLoginRequest>,
 ) -> Result<Json<AdminLoginResponse>, AppError> {
@@ -283,29 +488,63 @@ pub async fn admin_login(
         ));
     }
 
-    // Throttle password attempts. The admin password is bcrypt-hashed (slow to
-    // verify) but with no IP-level limit a determined attacker can still mount
-    // a long-running guess campaign. 5 attempts / minute / IP is plenty for
-    // honest typos.
-    let ip = client_ip(&headers, "unknown");
+    // Throttling here is in two parts, and the ORDER is the whole point.
+    //
+    // A single tight IP-keyed bucket checked before the password was verified made this
+    // endpoint a denial-of-service against its own operator. Every guest at the venue shares
+    // one public IP behind NAT, `/admin/login` is a public linkable page, and the check ran
+    // BEFORE `verify_password` — so five requests a minute from any phone in the room kept the
+    // bucket permanently full and the admin, on that same IP, could never spend a slot.
+    // Successful logins consumed budget too, so a typo plus a retry on two devices did it by
+    // accident. And the escape hatch was circular: `admin_login_rate_enabled` can only be
+    // flipped through `PATCH /admin/config`, which needs the session being blocked.
+    let ip = client_ip(&headers, &peer.ip().to_string());
     let rate_limits_on = config::get_bool(&state.config_cache, "rate_limits_enabled", true).await;
     let admin_rate_on =
         config::get_bool(&state.config_cache, "admin_login_rate_enabled", true).await;
+
+    // Part 1: a deliberately GENEROUS ceiling whose only job is to bound bcrypt CPU. Cost-12
+    // verification is ~250 ms of a core, so an unbounded endpoint is a CPU exhaustion vector
+    // regardless of whether anyone guesses right. No human typing a password reaches this.
     if rate_limits_on
         && admin_rate_on
-        && !state
-            .rate_limiter
-            .check(format!("admin_login:{ip}"), 5, Duration::from_secs(60))
+        && let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
+            format!("admin_login_cpu:{ip}"),
+            ADMIN_LOGIN_CPU_CEILING,
+            Duration::from_secs(60),
+        )
     {
         return Err(AppError::TooManyRequests(
             "Zu viele Anmeldeversuche. Bitte warte kurz und versuche es erneut.".into(),
-            None,
+            Some(retry_after_secs),
         ));
     }
 
-    let valid = bcrypt::verify(&body.password, &state.config.admin_password_hash).unwrap_or(false);
+    let valid = verify_password(
+        body.password.clone(),
+        state.config.admin_password_hash.clone(),
+    )
+    .await;
 
     if !valid {
+        // Part 2: the tight bucket, charged ONLY on a wrong password. A correct password is
+        // never rate-limited, so no amount of guessing by anyone else can lock the operator
+        // out — which also dissolves the circular escape hatch above. Brute force is still
+        // bounded: every wrong guess costs a slot, and slots are per-IP.
+        if rate_limits_on
+            && admin_rate_on
+            && let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
+                format!("admin_login_fail:{ip}"),
+                5,
+                Duration::from_secs(60),
+            )
+        {
+            tracing::warn!(ip = %ip, "admin_login: wrong password, failure bucket exhausted");
+            return Err(AppError::TooManyRequests(
+                "Zu viele Anmeldeversuche. Bitte warte kurz und versuche es erneut.".into(),
+                Some(retry_after_secs),
+            ));
+        }
         tracing::warn!(ip = %ip, "admin_login: wrong password");
         return Err(AppError::Unauthorized("Falsches Passwort.".into()));
     }
@@ -317,28 +556,16 @@ pub async fn admin_login(
     )
     .await?;
 
-    // Find or create the admin user for this event
-    let admin_name = "Admin";
-    let users = User::find_by_event_and_name(&state.pool, event.id, admin_name).await?;
-    let admin_user = if let Some(u) = users.into_iter().find(|u| u.role == UserRole::Admin) {
-        u
-    } else {
-        // Admin authenticates via password, but the schema still requires a PIN
-        // hash. Generate a random unguessable PIN so the recovery path remains
-        // unusable as an escalation route even if the role flag ever got cleared.
-        let dummy_pin: String = (0..32)
-            .map(|_| rand::rng().random_range(b'a'..=b'z') as char)
-            .collect();
-        let dummy_hash =
-            bcrypt::hash(&dummy_pin, 4).map_err(|e| AppError::Internal(anyhow::anyhow!(e)))?;
-        let user = User::create(&state.pool, event.id, admin_name, &dummy_hash).await?;
-        sqlx::query("UPDATE \"user\" SET role = 'admin' WHERE id = $1")
-            .bind(user.id)
-            .execute(&state.pool)
-            .await?;
-        User::find_by_id(&state.pool, user.id)
-            .await?
-            .ok_or_else(|| AppError::Internal(anyhow::anyhow!("admin user creation failed")))?
+    // Find or create the admin user for this event — BY ROLE, never by name.
+    //
+    // The name lookup this replaces is what made admin login brickable. Migration 007 makes
+    // display_name unique per event case-insensitively and `join` had no reserved-name guard,
+    // so a guest joining as "admin" before the operator's first login made the lookup miss on
+    // role, the fallback `create("Admin")` violate that index, and `?` return a permanent 500 —
+    // taking out moderation, config and gallery release with no in-app recovery.
+    let admin_user = match User::find_admin_for_event(&state.pool, event.id).await? {
+        Some(u) => u,
+        None => create_admin_user(&state, event.id).await?,
     };
 
     tracing::info!(user_id = %admin_user.id, event_id = %event.id, ip = %ip, "admin_login: success");
@@ -363,6 +590,51 @@ pub async fn admin_login(
     }))
 }
 
+/// Create this event's admin row on first successful admin login.
+///
+/// Prefers the name "Admin". If a legacy database has a guest squatting on it — the state
+/// migration 023 renames away, but a row could also predate that or be created between
+/// migrations — falls back to a suffixed name rather than failing the login.
+///
+/// PROMOTING THE SQUATTING ROW WOULD BE A SERIOUS MISTAKE, and is the obvious-looking fix, so
+/// it is spelled out: that row carries a `recovery_pin_hash` the guest knows. Setting
+/// `role = 'admin'` on it would hand them the admin dashboard through `/recover`, permanently,
+/// via a path that needs no password. A separate row under an uglier name is worse UX and much
+/// better security — and since the lookup is now by role, the fallback name never has to be
+/// guessed again on a later login.
+async fn create_admin_user(state: &AppState, event_id: Uuid) -> Result<User, AppError> {
+    // Admin authenticates via password, but the schema still requires a PIN hash. Generate a
+    // random unguessable one so the recovery path stays unusable as an escalation route even if
+    // the role flag were ever cleared.
+    let dummy_pin: String = (0..32)
+        .map(|_| rand::rng().random_range(b'a'..=b'z') as char)
+        .collect();
+    let dummy_hash = hash_password(dummy_pin, 4).await?;
+
+    match User::create_with_role(&state.pool, event_id, "Admin", &dummy_hash, UserRole::Admin).await
+    {
+        Ok(u) => Ok(u),
+        Err(sqlx::Error::Database(db)) if db.is_unique_violation() => {
+            let fallback = format!("Admin-{}", &Uuid::new_v4().to_string()[..8]);
+            tracing::warn!(
+                %event_id, %fallback,
+                "the name \"Admin\" is held by a non-admin user; creating the admin under a \
+                 fallback name. Rename that guest to free it — do NOT promote their row, they \
+                 know its recovery PIN."
+            );
+            Ok(User::create_with_role(
+                &state.pool,
+                event_id,
+                &fallback,
+                &dummy_hash,
+                UserRole::Admin,
+            )
+            .await?)
+        }
+        Err(e) => Err(e.into()),
+    }
+}
+
 pub async fn logout(State(state): State<AppState>, auth: AuthUser) -> Result<StatusCode, AppError> {
     Session::delete_by_token_hash(&state.pool, &auth.token_hash).await?;
     Ok(StatusCode::NO_CONTENT)
@@ -390,28 +662,55 @@ pub struct PinResetRequestBody {
 /// feed already exposes.
 pub async fn request_pin_reset(
     State(state): State<AppState>,
+    ConnectInfo(peer): ConnectInfo<SocketAddr>,
     headers: HeaderMap,
     Json(body): Json<PinResetRequestBody>,
 ) -> Result<StatusCode, AppError> {
-    let display_name = body.display_name.trim();
-    let ip = client_ip(&headers, "unknown");
+    let ip = client_ip(&headers, &peer.ip().to_string());
     let rate_limits_on = config::get_bool(&state.config_cache, "rate_limits_enabled", true).await;
+
+    // Coarse per-IP ceiling FIRST, keyed only on the IP so its key is bounded by construction.
+    // /join got one of these in migration 017 and /recover in 019; this third unauthenticated
+    // endpoint was simply missed — migration 019's own comment describes exactly this attack.
+    // Without it, the per-name bucket below is no ceiling at all: the name is attacker-chosen,
+    // so cycling names mints a fresh bucket every request.
+    if rate_limits_on {
+        let ip_ceiling =
+            config::get_usize(&state.config_cache, "pin_reset_ip_rate_per_min", 30).await;
+        if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
+            format!("pin_reset_ip:{ip}"),
+            ip_ceiling,
+            Duration::from_secs(60),
+        ) {
+            return Err(AppError::TooManyRequests(
+                "Zu viele Anfragen. Bitte warte kurz und versuche es erneut.".into(),
+                Some(retry_after_secs),
+            ));
+        }
+    }
+
+    // Validated BEFORE the per-name key is built, so an unbounded name can never be retained in
+    // the limiter map. NOTE the 204: this endpoint's contract is that it answers identically
+    // whether or not the name exists, so it cannot enumerate guests. A 400 here would be a new
+    // signal — it would distinguish a malformed name from a well-formed unknown one. Silence is
+    // the correct response, and matches what an empty name already did.
+    let Ok(display_name) = validate_display_name(&body.display_name) else {
+        return Ok(StatusCode::NO_CONTENT);
+    };
+
     if rate_limits_on {
         let name_key = display_name.to_lowercase();
-        if !state.rate_limiter.check(
+        if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
             format!("pin_reset_req:{ip}:{name_key}"),
             3,
             Duration::from_secs(15 * 60),
         ) {
             return Err(AppError::TooManyRequests(
                 "Zu viele Anfragen. Bitte warte kurz und versuche es erneut.".into(),
-                None,
+                Some(retry_after_secs),
             ));
         }
     }
-    if display_name.is_empty() {
-        return Ok(StatusCode::NO_CONTENT);
-    }
 
     // Single statement so the existing-name and unknown-name paths do IDENTICAL work
     // (same event+user index scans, an INSERT that matches 0 rows for an unknown name) —
@@ -441,3 +740,79 @@ pub async fn request_pin_reset(
 
     Ok(StatusCode::NO_CONTENT)
 }
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Control characters are rejected at the door. Newlines in particular: several 4xx messages
+    /// interpolate the display name and those are logged, so a name carrying a newline plus a
+    /// plausible prefix would let two unauthenticated requests forge lines in the event's only
+    /// forensic record. `error.rs` escapes on output too; this keeps it out of the database and
+    /// the feed byline in the first place.
+    #[test]
+    fn a_display_name_may_not_carry_control_characters() {
+        for bad in ["Anna\nERROR forged", "Anna\rX", "Anna\u{0}X", "A\u{7}B"] {
+            assert!(validate_display_name(bad).is_err(), "{bad:?} must be rejected");
+        }
+        // Real guests have accents, emoji and non-Latin names — never reject those.
+        for good in ["Anna", "Zo\u{eb}", "Jos\u{e9}", "\u{5c71}\u{7530}", "Anna \u{1f389}"] {
+            assert!(validate_display_name(good).is_ok(), "{good:?} must be allowed");
+        }
+    }
+
+    /// THE defect, stated as arithmetic: the account-lock threshold sat BELOW the per-(IP, name)
+    /// attempt ceiling, so a single IP could exhaust it and lock any guest whose display name is
+    /// visible on the feed — every 15 minutes, indefinitely. The tier meant to protect a guest
+    /// was the cheapest way to attack them.
+    ///
+    /// The fix is the ORDERING, not either number on its own, so that is what this pins.
+    #[test]
+    fn one_ip_cannot_reach_the_account_lock() {
+        assert!(
+            PIN_LOCK_THRESHOLD as usize >= RECOVER_NAME_CEILING_MAX * 3,
+            "locking a victim must require at least three distinct sources; \
+             threshold {PIN_LOCK_THRESHOLD} vs the ENFORCED per-IP ceiling \
+             {RECOVER_NAME_CEILING_MAX} (the default is {RECOVER_NAME_CEILING_DEFAULT})"
+        );
+    }
+
+    /// Raising the threshold must not quietly weaken brute-force resistance. 4-digit PINs, and
+    /// the lockout window is 15 minutes, so an attacker gets PIN_LOCK_THRESHOLD tries per window.
+    #[test]
+    fn the_raised_threshold_still_makes_guessing_a_four_digit_pin_impractical() {
+        let attempts_per_hour = PIN_LOCK_THRESHOLD as u64 * 4; // four 15-minute windows
+        let hours_for_full_keyspace = 10_000 / attempts_per_hour;
+        assert!(
+            hours_for_full_keyspace >= 168,
+            "exhausting 10k PINs would take {hours_for_full_keyspace}h — under a week is too fast"
+        );
+    }
+
+    #[test]
+    fn reserved_names_are_matched_case_insensitively_and_trimmed() {
+        for name in ["admin", "Admin", "ADMIN", "  Host  ", "EventSnap"] {
+            assert!(is_reserved_display_name(name), "{name} must be reserved");
+        }
+    }
+
+    /// A substring match here would reject perfectly ordinary names, which is a worse outcome
+    /// than the impersonation the list guards against.
+    #[test]
+    fn names_that_merely_contain_a_reserved_word_are_allowed() {
+        for name in ["Administrata", "Hostess", "Adminah", "Ghost", "hosting"] {
+            assert!(!is_reserved_display_name(name), "{name} must be allowed");
+        }
+    }
+
+    #[test]
+    fn display_names_are_bounded_before_they_can_become_a_rate_limit_key() {
+        assert!(validate_display_name("  Lena  ").is_ok());
+        assert_eq!(validate_display_name("  Lena  ").unwrap(), "Lena");
+        // The case that made the limiter itself the exhaustion primitive.
+        assert!(validate_display_name(&"a".repeat(51)).is_err());
+        assert!(validate_display_name(&"a".repeat(2_000_000)).is_err());
+        assert!(validate_display_name("   ").is_err());
+        assert!(validate_display_name("bad\0name").is_err());
+    }
+}

backend/src/config.rs

@@ -17,24 +17,92 @@ fn looks_placeholder(s: &str) -> bool {
         || lower.contains("placeholder")
 }
 
+/// A bcrypt hash is exactly 60 characters and opens with `$2<variant>$<cost>$`.
+///
+/// Checking the SHAPE, not just placeholder-ness, is what catches a hash silently mangled in
+/// transit. The `$` segments are variable-expansion bait for both shell quoting and Docker
+/// Compose's `env_file` parsing, and a mangled hash is not a placeholder — so without this it
+/// passes every other guard here, the app boots green, `/health` returns `ok`, and every admin
+/// login 401s.
+///
+/// That failure is unrecoverable mid-event, which is why it is worth a hard fail at boot: the
+/// Admin row is created BY a successful admin login (`auth/handlers.rs`), and only an Admin or
+/// Host can promote a Host. No admin login therefore means no host at all — the event cannot be
+/// closed, the gallery cannot be released, and nothing can be moderated.
+fn looks_bcrypt(s: &str) -> bool {
+    let b = s.as_bytes();
+    s.len() == 60
+        && b[0] == b'$'
+        && b[1] == b'2'
+        && matches!(b[2], b'a' | b'b' | b'x' | b'y')
+        && b[3] == b'$'
+        && b[4].is_ascii_digit()
+        && b[5].is_ascii_digit()
+        && b[6] == b'$'
+}
+
 /// Enforce secret hygiene. In production every guard is hard-fail: a booting app
 /// with a publicly-known signing key is worse than one that refuses to start.
 /// Outside production the dev sentinel is tolerated (warned) so local dev is frictionless.
-fn validate_secrets(is_prod: bool, jwt_secret: &str, admin_password_hash: &str) -> Result<()> {
+///
+/// EVERY failure is collected and reported together. Returning on the first one made fixing two
+/// secrets cost two boot cycles — the operator rotates JWT_SECRET, restarts, and only then learns
+/// about ADMIN_PASSWORD_HASH. Restarting this stack is not free (Caddy waits on the unhealthy app),
+/// and each avoidable cycle is another chance to reach for `down -v`.
+fn validate_secrets(
+    is_prod: bool,
+    jwt_secret: &str,
+    admin_password_hash: &str,
+    database_url: &str,
+) -> Result<()> {
     if is_prod {
+        let mut problems: Vec<&str> = Vec::new();
         if looks_placeholder(jwt_secret) {
-            return Err(anyhow!(
-                "Refusing to start in production with a placeholder JWT_SECRET — \
-                 rotate it (openssl rand -hex 64)."
-            ));
-        }
-        if jwt_secret.len() < 32 {
-            return Err(anyhow!("JWT_SECRET must be at least 32 characters."));
+            problems.push(
+                "JWT_SECRET is still the .env.example placeholder — rotate it \
+                 (openssl rand -hex 64).",
+            );
+        } else if jwt_secret.len() < 32 {
+            problems.push("JWT_SECRET must be at least 32 characters.");
         }
         if admin_password_hash.is_empty() || looks_placeholder(admin_password_hash) {
+            problems.push(
+                "ADMIN_PASSWORD_HASH is unset or still the .env.example placeholder — generate one \
+                 (docker run --rm caddy:2-alpine caddy hash-password --plaintext '<password>').",
+            );
+        } else if !looks_bcrypt(admin_password_hash) {
+            problems.push(
+                "ADMIN_PASSWORD_HASH is not a well-formed bcrypt hash: expected exactly 60 \
+                 characters starting `$2b$12$…`. First look at the value that actually reached \
+                 the app — `docker compose exec app printenv ADMIN_PASSWORD_HASH` — and compare \
+                 it to .env character for character. In .env, SINGLE-QUOTE the hash \
+                 ('$2b$12$…'): Compose uses single-quoted env_file values literally, so the `$` \
+                 segments survive. Double them to `$$` ONLY when setting the value under \
+                 `environment:` in docker-compose.yml — doing that in .env corrupts a hash that \
+                 would otherwise have worked. Regenerate with: \
+                 docker run --rm caddy:2-alpine caddy hash-password --plaintext '<password>'",
+            );
+        }
+        // The DATABASE_URL carries the Postgres password, so a placeholder here means the stack is
+        // running on `CHANGE_ME_use_a_strong_password` — a credential published in the repo. The
+        // app used to boot green on it, because this guard only ever covered the two secrets it
+        // was written for and nothing else looked at POSTGRES_PASSWORD at all.
+        //
+        // Read POSTGRES_PASSWORD's docs before changing this: it is applied ONLY at initdb, so the
+        // remedy is not "edit .env and restart" — see the 28P01 diagnostic in db.rs.
+        if looks_placeholder(database_url) {
+            problems.push(
+                "DATABASE_URL still carries the .env.example placeholder password — set a strong \
+                 one (openssl rand -hex 24) in BOTH DATABASE_URL and POSTGRES_PASSWORD.",
+            );
+        }
+        if !problems.is_empty() {
             return Err(anyhow!(
-                "Refusing to start in production without a real ADMIN_PASSWORD_HASH — \
-                 generate one (htpasswd -bnBC 12 '' <password> | tr -d ':\\n')."
+                "Refusing to start in production — {} secret(s) still unset or placeholder:\n  - {}\n\
+                 ALL secrets must be set BEFORE the first `docker compose up -d`: Postgres bakes \
+                 POSTGRES_PASSWORD into its data directory on first boot and ignores later changes.",
+                problems.len(),
+                problems.join("\n  - ")
             ));
         }
     } else if jwt_secret == DEV_JWT_SECRET_SENTINEL {
@@ -63,8 +131,25 @@ pub struct AppConfig {
     pub app_port: u16,
     /// Number of concurrent media compression workers (read once at boot).
     pub compression_concurrency: usize,
+    /// Master switch for the comment feature (env `COMMENTS_ENABLED`, default true).
+    /// When false the backend rejects new comments and the frontend hides the whole
+    /// comment UI. Existing comments stay in the DB (hidden), so flipping it back
+    /// restores them. Boot-time immutable, like `compression_concurrency`.
+    pub comments_enabled: bool,
+    /// Default colour theme, used as the fallback when the DB config keys are unset.
+    /// Runtime overrides live in the `config` table (admin UI); these env vars only
+    /// seed the initial default. `preset` is an id the frontend knows (e.g.
+    /// "champagne-gold", "rose", … or "custom"); the two seeds are `#rrggbb` brand +
+    /// accent colours the whole palette is derived from.
+    pub default_theme_preset: String,
+    pub default_theme_primary: String,
+    pub default_theme_accent: String,
 }
 
+/// The shipped default brand/accent seed (champagne gold — matches the hand-tuned
+/// ramp in tailwind-theme.css). Kept here so an unset env still yields the current look.
+const DEFAULT_THEME_SEED: &str = "#8a6a2b";
+
 impl AppConfig {
     pub fn from_env() -> Result<Self> {
         let app_env = std::env::var("APP_ENV").unwrap_or_else(|_| "development".to_string());
@@ -72,11 +157,12 @@ impl AppConfig {
 
         let jwt_secret = std::env::var("JWT_SECRET").context("JWT_SECRET must be set")?;
         let admin_password_hash = std::env::var("ADMIN_PASSWORD_HASH").unwrap_or_default();
+        let database_url = std::env::var("DATABASE_URL").context("DATABASE_URL must be set")?;
 
-        validate_secrets(is_prod, &jwt_secret, &admin_password_hash)?;
+        validate_secrets(is_prod, &jwt_secret, &admin_password_hash, &database_url)?;
 
         Ok(Self {
-            database_url: std::env::var("DATABASE_URL").context("DATABASE_URL must be set")?,
+            database_url,
             jwt_secret,
             session_expiry_days: std::env::var("SESSION_EXPIRY_DAYS")
                 .unwrap_or_else(|_| "30".to_string())
@@ -100,6 +186,20 @@ impl AppConfig {
                 .and_then(|v| v.parse().ok())
                 .filter(|&n| n >= 1)
                 .unwrap_or(2),
+            comments_enabled: std::env::var("COMMENTS_ENABLED")
+                .map(|v| {
+                    !matches!(
+                        v.trim().to_ascii_lowercase().as_str(),
+                        "false" | "0" | "no" | "off"
+                    )
+                })
+                .unwrap_or(true),
+            default_theme_preset: std::env::var("THEME_PRESET")
+                .unwrap_or_else(|_| "champagne-gold".to_string()),
+            default_theme_primary: std::env::var("THEME_PRIMARY")
+                .unwrap_or_else(|_| DEFAULT_THEME_SEED.to_string()),
+            default_theme_accent: std::env::var("THEME_ACCENT")
+                .unwrap_or_else(|_| DEFAULT_THEME_SEED.to_string()),
         })
     }
 }
@@ -109,13 +209,22 @@ mod tests {
     use super::*;
 
     const REAL_SECRET: &str = "a1b2c3d4e5f6a7b8c9d0e1f2a3b4c5d6a7b8c9d0e1f2a3b4c5d6e7f8a9b0c1d2";
-    const REAL_HASH: &str = "$2y$12$abcdefghijklmnopqrstuv.wxyzABCDEFGHIJKLMNOPQRSTUVWXYZ012";
+    // A structurally valid bcrypt hash: exactly 60 chars, `$2y$12$` + 53 of salt/digest.
+    // The shape matters — `looks_bcrypt` enforces it, so a fixture of the wrong length
+    // would assert the opposite of what these tests claim.
+    const REAL_HASH: &str = "$2y$12$abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXY01";
+    const REAL_DB_URL: &str = "postgres://eventsnap:7f3a9c1e5b2d8a4f@db:5432/eventsnap";
 
     #[test]
     fn prod_rejects_shipped_placeholder_secret() {
         // The exact string shipped in `.env` — >32 chars, so it must be caught by
         // the substring guard, not the length check.
-        let err = validate_secrets(true, "change_me_to_a_random_64_byte_hex_string", REAL_HASH);
+        let err = validate_secrets(
+            true,
+            "change_me_to_a_random_64_byte_hex_string",
+            REAL_HASH,
+            REAL_DB_URL,
+        );
         assert!(
             err.is_err(),
             "placeholder JWT_SECRET must be rejected in prod"
@@ -124,29 +233,133 @@ mod tests {
 
     #[test]
     fn prod_rejects_dev_sentinel_and_short_secret() {
-        assert!(validate_secrets(true, DEV_JWT_SECRET_SENTINEL, REAL_HASH).is_err());
-        assert!(validate_secrets(true, "tooshort", REAL_HASH).is_err());
+        assert!(validate_secrets(true, DEV_JWT_SECRET_SENTINEL, REAL_HASH, REAL_DB_URL).is_err());
+        assert!(validate_secrets(true, "tooshort", REAL_HASH, REAL_DB_URL).is_err());
+    }
+
+    /// The failure this guards is silent and unrecoverable mid-event: a hash whose `$`
+    /// segments were eaten by shell or Compose interpolation is NOT a placeholder, so every
+    /// other guard passes, the app boots green and `/health` reports ok — and then every
+    /// admin login 401s, which (because the Admin row is created by a successful login, and
+    /// only an Admin/Host can promote a Host) means no host exists for the whole event.
+    #[test]
+    fn prod_rejects_mangled_admin_hash() {
+        // What `$2y$12$…` degrades to once `$2y`/`$12` are read as unset variables.
+        assert!(validate_secrets(true, REAL_SECRET, "abcdefghijklmnop", REAL_DB_URL).is_err());
+        // Right prefix, truncated body — still not a usable hash.
+        assert!(validate_secrets(true, REAL_SECRET, "$2y$12$tooshort", REAL_DB_URL).is_err());
+        // Correct length but no bcrypt prefix at all.
+        assert!(validate_secrets(true, REAL_SECRET, &"x".repeat(60), REAL_DB_URL).is_err());
+        // All three shipped bcrypt variants stay acceptable.
+        let body = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXY01";
+        for variant in ["2a", "2b", "2y"] {
+            let hash = format!("${variant}$12${body}");
+            assert_eq!(hash.len(), 60);
+            assert!(
+                validate_secrets(true, REAL_SECRET, &hash, REAL_DB_URL).is_ok(),
+                "bcrypt variant ${variant}$ must be accepted"
+            );
+        }
     }
 
     #[test]
     fn prod_rejects_missing_or_placeholder_admin_hash() {
-        assert!(validate_secrets(true, REAL_SECRET, "").is_err());
-        assert!(validate_secrets(true, REAL_SECRET, "$2y$12$placeholder_replace_me").is_err());
+        assert!(validate_secrets(true, REAL_SECRET, "", REAL_DB_URL).is_err());
+        assert!(
+            validate_secrets(
+                true,
+                REAL_SECRET,
+                "$2y$12$placeholder_replace_me",
+                REAL_DB_URL
+            )
+            .is_err()
+        );
+    }
+
+    /// The stack used to come up GREEN on the database password published in the repo: this guard
+    /// covered the two secrets it was written for, and nothing anywhere looked at the Postgres
+    /// credential. README step 2 doesn't name POSTGRES_PASSWORD either, so following the
+    /// documented procedure verbatim shipped it.
+    #[test]
+    fn prod_rejects_the_shipped_placeholder_database_password() {
+        let shipped = "postgres://eventsnap:CHANGE_ME_use_a_strong_password@db:5432/eventsnap";
+        let err = validate_secrets(true, REAL_SECRET, REAL_HASH, shipped).unwrap_err();
+        assert!(
+            err.to_string().contains("DATABASE_URL"),
+            "the refusal must name DATABASE_URL, not just fail: {err}"
+        );
+        // And it must point at the initdb trap, or the operator edits .env, restarts, and lands
+        // in a permanent auth-failure loop instead.
+        assert!(
+            err.to_string().contains("POSTGRES_PASSWORD"),
+            "the refusal must name POSTGRES_PASSWORD as the other half: {err}"
+        );
+    }
+
+    /// Every problem in ONE message. Reporting them one per boot made fixing two secrets cost two
+    /// restart cycles, on a stack where Caddy waits on the unhealthy app the whole time.
+    #[test]
+    fn prod_reports_every_placeholder_at_once() {
+        let err = validate_secrets(
+            true,
+            "change_me_to_a_random_64_byte_hex_string",
+            "$2y$12$placeholder_replace_me",
+            "postgres://eventsnap:CHANGE_ME_use_a_strong_password@db:5432/eventsnap",
+        )
+        .unwrap_err()
+        .to_string();
+        for expected in ["JWT_SECRET", "ADMIN_PASSWORD_HASH", "DATABASE_URL"] {
+            assert!(err.contains(expected), "{expected} missing from: {err}");
+        }
+        assert!(
+            err.contains("3 secret(s)"),
+            "the count must match what is listed: {err}"
+        );
     }
 
     #[test]
     fn prod_accepts_real_secrets() {
-        assert!(validate_secrets(true, REAL_SECRET, REAL_HASH).is_ok());
+        assert!(validate_secrets(true, REAL_SECRET, REAL_HASH, REAL_DB_URL).is_ok());
+    }
+
+    /// A real password that happens to contain no placeholder substring must pass — including one
+    /// with URL-ish punctuation, so the guard can't be mistaken for a URL validator.
+    #[test]
+    fn prod_accepts_a_real_database_url_with_awkward_punctuation() {
+        assert!(
+            validate_secrets(
+                true,
+                REAL_SECRET,
+                REAL_HASH,
+                "postgres://eventsnap:aB3%24xY9-_.qW@db:5432/eventsnap"
+            )
+            .is_ok()
+        );
+    }
+
+    /// The e2e stack runs without APP_ENV=production, so none of this applies there — but assert
+    /// it, because a guard that tripped in e2e would be found the hard way.
+    #[test]
+    fn non_prod_ignores_a_placeholder_database_url() {
+        assert!(
+            validate_secrets(
+                false,
+                REAL_SECRET,
+                "",
+                "postgres://eventsnap:CHANGE_ME_use_a_strong_password@db:5432/eventsnap"
+            )
+            .is_ok()
+        );
     }
 
     #[test]
     fn non_prod_tolerates_dev_sentinel() {
-        assert!(validate_secrets(false, DEV_JWT_SECRET_SENTINEL, "").is_ok());
+        assert!(validate_secrets(false, DEV_JWT_SECRET_SENTINEL, "", REAL_DB_URL).is_ok());
     }
 
     #[test]
     fn non_prod_still_rejects_short_non_sentinel_secret() {
-        assert!(validate_secrets(false, "tooshort", "").is_err());
+        assert!(validate_secrets(false, "tooshort", "", REAL_DB_URL).is_err());
     }
 
     #[test]
@@ -154,9 +367,23 @@ mod tests {
         // looks_placeholder lowercases before matching — an upper/mixed-case
         // placeholder must still be rejected in prod.
         assert!(
-            validate_secrets(true, "CHANGE_ME_TO_A_RANDOM_64_BYTE_HEX_STRING", REAL_HASH).is_err()
+            validate_secrets(
+                true,
+                "CHANGE_ME_TO_A_RANDOM_64_BYTE_HEX_STRING",
+                REAL_HASH,
+                REAL_DB_URL
+            )
+            .is_err()
+        );
+        assert!(
+            validate_secrets(
+                true,
+                REAL_SECRET,
+                "$2Y$12$PLACEHOLDER_replace_me",
+                REAL_DB_URL
+            )
+            .is_err()
         );
-        assert!(validate_secrets(true, REAL_SECRET, "$2Y$12$PLACEHOLDER_replace_me").is_err());
     }
 
     #[test]
@@ -166,7 +393,7 @@ mod tests {
         const LEN_31: &str = "abcdefghijklmnopqrstuvwxyz01234";
         assert_eq!(LEN_32.len(), 32);
         assert_eq!(LEN_31.len(), 31);
-        assert!(validate_secrets(true, LEN_32, REAL_HASH).is_ok());
-        assert!(validate_secrets(true, LEN_31, REAL_HASH).is_err());
+        assert!(validate_secrets(true, LEN_32, REAL_HASH, REAL_DB_URL).is_ok());
+        assert!(validate_secrets(true, LEN_31, REAL_HASH, REAL_DB_URL).is_err());
     }
 }

backend/src/db.rs

@@ -4,17 +4,90 @@ use sqlx::postgres::PgPoolOptions;
 
 const DEFAULT_MAX_CONNECTIONS: u32 = 10;
 
+/// SQLSTATE for `invalid_password`.
+const PG_INVALID_PASSWORD: &str = "28P01";
+
+/// Turn the one connect failure with an unguessable cause into a self-explaining one.
+///
+/// `POSTGRES_PASSWORD` is honoured ONLY when Postgres initialises its data directory. Change it in
+/// `.env` afterwards and the app authenticates with the new password against a volume that still
+/// holds the old one — a permanent restart loop whose only symptom is
+/// `password authentication failed`.
+///
+/// The production secret guard makes that sequence NEARLY CERTAIN rather than rare: it stops the
+/// app on the first `docker compose up -d`, but not the `db` service in that same command, which
+/// initialises and bakes in whatever password was in `.env` at that moment. So the intended
+/// recovery — see the refusal, fix your secrets, boot again — is exactly the sequence that breaks
+/// it. Nothing in the error names the cause, and the remedy destroys data, so it is the last thing
+/// an operator should guess at.
+fn explain_auth_failure(err: &sqlx::Error) {
+    let is_auth_failure = match err {
+        sqlx::Error::Database(db) => db.code().as_deref() == Some(PG_INVALID_PASSWORD),
+        _ => false,
+    };
+    if !is_auth_failure {
+        return;
+    }
+    tracing::error!(
+        "Postgres rejected the credentials in DATABASE_URL (SQLSTATE {PG_INVALID_PASSWORD}).\n\
+         \n\
+         This almost always means POSTGRES_PASSWORD was changed AFTER the database volume was \
+         first created. Postgres applies that variable only when it initialises its data \
+         directory; editing .env and restarting does not change the stored password, so the two \
+         drift apart permanently.\n\
+         \n\
+         If the event has NOT started and you have no data worth keeping:\n\n    \
+         docker compose down -v && docker compose up -d\n\n\
+         (-v DELETES the database, the uploaded media and the exports. There is no undo.)\n\
+         \n\
+         If you DO have data: restore the old password into DATABASE_URL instead, or change the \
+         stored one with ALTER ROLE inside the running db container. Never reach for -v to fix a \
+         login problem on a live event."
+    );
+}
+
 pub async fn create_pool(database_url: &str) -> Result<PgPool> {
     let max_connections = std::env::var("DATABASE_MAX_CONNECTIONS")
         .ok()
         .and_then(|s| s.parse::<u32>().ok())
         .unwrap_or(DEFAULT_MAX_CONNECTIONS);
 
-    let pool = PgPoolOptions::new()
+    let pool = match PgPoolOptions::new()
         .max_connections(max_connections)
+        // Fail fast instead of parking. sqlx's default is 30s, which on a DB blip means every
+        // request AND all ~100 SSE session revalidations sit on the pool for half a minute
+        // before erroring — the app looks hung rather than degraded, and the backlog outlives
+        // the blip. Five seconds is far longer than a healthy acquire ever takes.
+        .acquire_timeout(std::time::Duration::from_secs(5))
+        // Keep a couple of connections warm so the first request after an idle stretch (the gap
+        // between setting the venue up and the guests arriving) doesn't pay TCP + auth.
+        .min_connections(2)
+        // Bound every statement server-side. Without this a single pathological query holds a
+        // pool slot indefinitely and no client-side timeout can take it back — the slot is only
+        // released when Postgres finishes. `lock_timeout` covers the same hazard for a row lock
+        // contended by, say, a release running against an in-flight upload.
+        .after_connect(|conn, _meta| {
+            Box::pin(async move {
+                // Two statements, two round-trips, deliberately. `sqlx::query` uses the extended
+                // query protocol, which permits exactly ONE statement per call — sending them as
+                // `SET a; SET b` makes every new connection fail, which surfaces as the pool
+                // never opening one at all and `create_pool` reporting a connect timeout.
+                sqlx::query("SET statement_timeout = '15s'")
+                    .execute(&mut *conn)
+                    .await?;
+                sqlx::query("SET lock_timeout = '5s'").execute(conn).await?;
+                Ok(())
+            })
+        })
         .connect(database_url)
         .await
-        .context("failed to connect to database")?;
+    {
+        Ok(pool) => pool,
+        Err(e) => {
+            explain_auth_failure(&e);
+            return Err(e).context("failed to connect to database");
+        }
+    };
 
     sqlx::migrate!()
         .run(&pool)

backend/src/error.rs

@@ -19,6 +19,12 @@ pub enum AppError {
     /// the client can treat it as *terminal* (413, no retry) instead of backing off and
     /// retrying a permanently-failing upload forever.
     QuotaExceeded(String),
+    /// The server is temporarily unable to serve this request — currently only pool
+    /// saturation. Distinct from `Internal` because it is TRANSIENT and the client should be
+    /// told so: a 500 reads as "this request is broken", while a 503 + Retry-After reads as
+    /// "come back shortly", which is what the upload queue's retry classifier needs to make
+    /// the right call. Second field: optional retry-after seconds.
+    ServiceUnavailable(String, Option<u64>),
     Internal(anyhow::Error),
 }
 
@@ -33,6 +39,9 @@ impl AppError {
             Self::Conflict(_) => (StatusCode::CONFLICT, "conflict"),
             Self::TooManyRequests(..) => (StatusCode::TOO_MANY_REQUESTS, "too_many_requests"),
             Self::QuotaExceeded(_) => (StatusCode::PAYLOAD_TOO_LARGE, "quota_exceeded"),
+            Self::ServiceUnavailable(..) => {
+                (StatusCode::SERVICE_UNAVAILABLE, "service_unavailable")
+            }
             Self::Internal(_) => (StatusCode::INTERNAL_SERVER_ERROR, "internal_error"),
         }
     }
@@ -46,6 +55,7 @@ impl AppError {
             | Self::NotFound(msg)
             | Self::Conflict(msg) => msg.clone(),
             Self::TooManyRequests(msg, _) => msg.clone(),
+            Self::ServiceUnavailable(msg, _) => msg.clone(),
             Self::QuotaExceeded(msg) => msg.clone(),
             Self::Internal(err) => {
                 tracing::error!("internal error: {err:#}");
@@ -58,13 +68,61 @@ impl AppError {
 impl IntoResponse for AppError {
     fn into_response(self) -> Response {
         let (status, code) = self.status_and_code();
-        let retry_after_secs = if let Self::TooManyRequests(_, Some(secs)) = &self {
-            Some(*secs)
-        } else {
-            None
+        // BOTH retry-carrying variants must be matched here. `message()` would fail to
+        // compile on a missing arm; this one would not — it would silently drop the header and
+        // the `retry_after_secs` body field, which is exactly the sort of omission that only
+        // shows up under the load the 503 exists for.
+        let retry_after_secs = match &self {
+            Self::TooManyRequests(_, secs) | Self::ServiceUnavailable(_, secs) => *secs,
+            _ => None,
         };
         let message = self.message();
 
+        // Log every 4xx. Until now they were invisible at ANY log level: tower_http's
+        // `ServerErrorsAsFailures` classifier counts a 4xx as a *success*, so it goes to
+        // `DefaultOnResponse` at DEBUG, and production runs at `info`. The consequence is that a
+        // misconfigured limit leaves no trace at all — if guests spend the evening hitting 429s
+        // on `upload_rate_per_hour`, or 413s on the storage quota, `docker compose logs` after
+        // the event contains nothing about it and the cause is unknowable.
+        //
+        // WARN rather than INFO because every variant here is a request that did not do what
+        // the guest asked. 5xx is excluded: `Internal` already logs with its full source chain
+        // in `message()` above, and the pool-exhaustion 503 logs at construction — logging again
+        // here would double every server-side failure.
+        //
+        // No request context is available: `into_response` receives only the error, so there is
+        // no path, method or user id to attach. Status + code + message is what can honestly be
+        // reported from here, and it is enough to see the SHAPE of a bad evening. Raising
+        // `tower_http` to DEBUG instead was considered and rejected — see the note in main.rs.
+        //
+        // `detail = ?message`, NOT `%message`. Two reasons, both learned the hard way:
+        //
+        //  * `message` is tracing's own reserved field for an event's format literal, so `%message`
+        //    printed unlabelled and would collide under a JSON layer.
+        //  * Debug formatting QUOTES AND ESCAPES the string, and several 4xx messages interpolate
+        //    attacker-chosen text — the guest's name in `Der Name "X" ist bereits vergeben.`, and
+        //    multipart/parse errors that echo their input. With Display formatting, a value
+        //    carrying a newline plus a plausible log prefix lets two unauthenticated requests
+        //    forge lines in the only forensic record an unattended event has.
+        //    `validate_display_name` now rejects control characters, so the name route is closed
+        //    at the source as well — but that is ONE input, and this line formats every 4xx
+        //    message in the app. Escaping here is what makes the guarantee general; do not
+        //    "simplify" it to `%message` on the grounds that names are already validated.
+        //
+        // 401 and 404 are logged at DEBUG rather than WARN. They carry no operator signal (an
+        // expired session, a mistyped URL) and they are the cheapest lines for a scanner to
+        // generate — at ~260 bytes each against the 30 MB the json-file driver retains
+        // (docker-compose.yml), a sustained flood could otherwise roll the whole window in
+        // minutes and destroy the post-event forensics this logging exists to provide.
+        if status.is_client_error() {
+            let noisy = status == StatusCode::UNAUTHORIZED || status == StatusCode::NOT_FOUND;
+            if noisy {
+                tracing::debug!(status = status.as_u16(), code, detail = ?message, "request rejected");
+            } else {
+                tracing::warn!(status = status.as_u16(), code, detail = ?message, "request rejected");
+            }
+        }
+
         let mut body = serde_json::json!({
             "error": code,
             "message": message,
@@ -93,6 +151,121 @@ impl From<anyhow::Error> for AppError {
 
 impl From<sqlx::Error> for AppError {
     fn from(err: sqlx::Error) -> Self {
-        Self::Internal(err.into())
+        match err {
+            // Pool saturation is load, not a bug. Reporting it as a 500 was actively harmful:
+            // the frontend's upload-queue classifier treats 5xx as transient and retries, so
+            // the retries piled straight back into the saturated pool with no Retry-After to
+            // pace them. A 503 says the same thing honestly and carries the backoff.
+            //
+            // `PoolClosed` stays `Internal` — it only happens during shutdown, where a 503
+            // would invite a retry against a server that is going away.
+            sqlx::Error::PoolTimedOut => {
+                tracing::warn!("database pool exhausted; shedding a request with 503");
+                Self::ServiceUnavailable(
+                    "Server ist gerade ausgelastet. Bitte versuche es in ein paar Sekunden erneut."
+                        .into(),
+                    Some(POOL_TIMEOUT_RETRY_AFTER_SECS),
+                )
+            }
+            other => Self::Internal(other.into()),
+        }
+    }
+}
+
+/// Retry-After for a shed request. Short: pool saturation clears in seconds once the queue
+/// drains, and a long value would make a brief spike feel like an outage.
+const POOL_TIMEOUT_RETRY_AFTER_SECS: u64 = 3;
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// `into_response` extracts `retry_after_secs` by MATCHING ON VARIANTS, so unlike
+    /// `message()` a missing arm is not a compile error — it silently drops the header. Pin the
+    /// behaviour for both retry-carrying variants.
+    #[test]
+    fn both_retry_carrying_variants_emit_retry_after() {
+        for err in [
+            AppError::TooManyRequests("slow down".into(), Some(42)),
+            AppError::ServiceUnavailable("busy".into(), Some(3)),
+        ] {
+            let expected = match &err {
+                AppError::TooManyRequests(_, Some(s)) | AppError::ServiceUnavailable(_, Some(s)) => {
+                    s.to_string()
+                }
+                _ => unreachable!(),
+            };
+            let resp = err.into_response();
+            assert_eq!(
+                resp.headers()
+                    .get(axum::http::header::RETRY_AFTER)
+                    .and_then(|v| v.to_str().ok()),
+                Some(expected.as_str()),
+                "a shed/throttled client must be told when to come back"
+            );
+        }
+    }
+
+    /// 4xx must be logged and 5xx must not be logged HERE — `Internal` logs its source chain in
+    /// `message()` and the pool-exhaustion 503 logs at construction, so a second line in
+    /// `into_response` would double every server-side failure in the post-event logs.
+    ///
+    /// The guard is `status.is_client_error()`, so this pins the classification rather than the
+    /// logging itself (which needs a subscriber to observe).
+    #[test]
+    fn only_client_errors_are_in_the_logged_band() {
+        for err in [
+            AppError::BadRequest("x".into()),
+            AppError::Unauthorized("x".into()),
+            AppError::Forbidden("x".into()),
+            AppError::UploadsLocked("x".into()),
+            AppError::NotFound("x".into()),
+            AppError::Conflict("x".into()),
+            AppError::TooManyRequests("x".into(), Some(1)),
+            AppError::QuotaExceeded("x".into()),
+        ] {
+            let (status, _) = err.status_and_code();
+            assert!(
+                status.is_client_error(),
+                "{status} should be in the 4xx band this logs"
+            );
+        }
+
+        for err in [
+            AppError::ServiceUnavailable("x".into(), Some(3)),
+            AppError::Internal(anyhow::anyhow!("boom")),
+        ] {
+            let (status, _) = err.status_and_code();
+            assert!(
+                !status.is_client_error(),
+                "{status} logs elsewhere; logging it here would double it"
+            );
+        }
+    }
+
+    /// Pool saturation is load, not a bug. A 500 makes the frontend's retry classifier pile
+    /// straight back into the saturated pool with no backoff to pace it.
+    #[test]
+    fn pool_exhaustion_sheds_with_503_but_shutdown_does_not() {
+        let shed: AppError = sqlx::Error::PoolTimedOut.into();
+        assert_eq!(
+            shed.status_and_code(),
+            (StatusCode::SERVICE_UNAVAILABLE, "service_unavailable")
+        );
+
+        // PoolClosed only happens during shutdown; a 503 there would invite a retry against a
+        // server that is going away.
+        let closing: AppError = sqlx::Error::PoolClosed.into();
+        assert_eq!(
+            closing.status_and_code(),
+            (StatusCode::INTERNAL_SERVER_ERROR, "internal_error")
+        );
+
+        // Everything else must keep its existing mapping.
+        let missing: AppError = sqlx::Error::RowNotFound.into();
+        assert_eq!(
+            missing.status_and_code(),
+            (StatusCode::INTERNAL_SERVER_ERROR, "internal_error")
+        );
     }
 }

backend/src/handlers/admin.rs

@@ -3,13 +3,14 @@ use std::time::Duration;
 
 use axum::Json;
 use axum::extract::{Query, State};
-use axum::http::{HeaderMap, StatusCode};
+use axum::http::StatusCode;
 use serde::{Deserialize, Serialize};
+use uuid::Uuid;
 
 use crate::auth::middleware::RequireAdmin;
 use crate::error::AppError;
 use crate::services::config;
-use crate::services::rate_limiter::client_ip;
+use crate::services::sse_tickets::TicketKind;
 use crate::state::AppState;
 
 // ── DTOs ─────────────────────────────────────────────────────────────────────
@@ -120,8 +121,28 @@ pub async fn patch_config(
         ("upload_rate_per_hour", true, 1.0, 100_000.0),
         ("feed_rate_per_min", true, 1.0, 100_000.0),
         ("export_rate_per_day", true, 1.0, 100_000.0),
+        // Loose per-IP ceiling on /join. The real anti-spam bucket is per (ip, name); this
+        // only bounds raw volume from one source, so it must stay well above the size of a
+        // party arriving at once (see migration 017).
+        ("join_ip_rate_per_min", true, 1.0, 100_000.0),
+        // Same shape for /recover: the per-(ip, name) bucket is the anti-guessing control,
+        // this only bounds a name-cycling flood in front of a cost-12 bcrypt (migration 019).
+        ("recover_ip_rate_per_min", true, 1.0, 100_000.0),
+        // Aggregate ceiling on likes + comments + comment deletions, per user per minute.
+        // These were the only mutating endpoints with no limit at all (migration 020).
+        ("social_rate_per_min", true, 1.0, 100_000.0),
         ("quota_tolerance", false, 0.0, 1.0),
         ("estimated_guest_count", true, 1.0, 1_000_000.0),
+        // The three limiters migration 025 introduced. All are READ at runtime
+        // (`upload.rs` for the edit limiter, `auth/handlers.rs` for the other two) and 025
+        // INSERTs all of them into `config`, so `GET /admin/config` listed them while
+        // `PATCH /admin/config` answered "Unbekannter Konfigurationsschlüssel" — the same
+        // dead-key defect the comment under BOOL_KEYS says was fixed for the two login
+        // toggles. These are precisely the knobs an operator reaches for while abuse is
+        // happening, which is the one moment a restart to change them is unaffordable.
+        ("upload_edit_rate_per_min", true, 1.0, 100_000.0),
+        ("recover_name_rate_per_15min", true, 1.0, 100_000.0),
+        ("pin_reset_ip_rate_per_min", true, 1.0, 100_000.0),
     ];
     const BOOL_KEYS: &[&str] = &[
         "rate_limits_enabled",
@@ -134,14 +155,35 @@ pub async fn patch_config(
         // missing from this allowlist — so the switch existed in code and could never be flipped.
         "admin_login_rate_enabled",
         "recover_rate_enabled",
+        "social_rate_enabled",
+        // Read by `upload::edit_upload`, inserted by migration 025, and until now unreachable
+        // from this endpoint — see the note in NUMERIC_SPECS.
+        "upload_edit_rate_enabled",
         "quota_enabled",
         "storage_quota_enabled",
         "upload_count_quota_enabled",
     ];
-    const TEXT_KEYS: &[&str] = &["privacy_note"];
+    const TEXT_KEYS: &[&str] = &[
+        "privacy_note",
+        "theme_preset",
+        "theme_primary",
+        "theme_accent",
+    ];
     const PRIVACY_NOTE_MAX_LEN: usize = 16 * 1024; // 16 KiB free text is plenty
+    // Preset ids the frontend knows how to render (mirror of PRESETS in
+    // frontend/src/lib/theme/palette.ts). "custom" means "use the theme_primary/accent
+    // seeds verbatim". Kept in sync by hand — a new preset must be added in both places.
+    const THEME_PRESETS: &[&str] = &[
+        "champagne-gold",
+        "rose",
+        "sage",
+        "dusk-blue",
+        "classic-silver",
+        "custom",
+    ];
 
     let mut privacy_note_changed = false;
+    let mut theme_changed = false;
 
     // Validate every key first so a bad value in the batch can't leave a partial
     // update behind — validation must fully precede any write.
@@ -169,6 +211,23 @@ pub async fn patch_config(
                     "Wert für {key} liegt außerhalb des zulässigen Bereichs ({min}–{max})."
                 )));
             }
+            // Zero is in range and catastrophic. `quota_tolerance` is the multiplier in
+            // `free_disk * tolerance / active_uploaders`, so 0 makes every per-user limit 0 and
+            // refuses EVERY upload — mid-event, with "Du hast dein Upload-Limit für dieses Event
+            // erreicht", an error naming the wrong cause entirely. `storage_quota_enabled` is the
+            // intended off-switch.
+            //
+            // Rejecting the value rather than raising the floor: very small tolerances are
+            // legitimate (they are how a large disk is throttled down to a sensible per-guest
+            // ceiling, and how the e2e quota tests steer it — around 1e-5 on a 174 GB volume), so
+            // a floor of, say, 0.01 would forbid real configurations to prevent one typo.
+            if key_str == "quota_tolerance" && n == 0.0 {
+                return Err(AppError::BadRequest(
+                    "quota_tolerance = 0 würde jeden Upload blockieren. Zum Abschalten der \
+                     Speicher-Quote stattdessen „Speicher-Quote aktiv“ ausschalten."
+                        .into(),
+                ));
+            }
         } else if BOOL_KEYS.contains(&key_str) {
             match value.trim().to_ascii_lowercase().as_str() {
                 "true" | "false" | "1" | "0" | "yes" | "no" | "on" | "off" => {}
@@ -186,8 +245,23 @@ pub async fn patch_config(
                     "Wert für {key} ist zu lang (max. {PRIVACY_NOTE_MAX_LEN} Zeichen)."
                 )));
             }
-            if key_str == "privacy_note" {
-                privacy_note_changed = true;
+            match key_str {
+                "privacy_note" => privacy_note_changed = true,
+                "theme_preset" => {
+                    if !THEME_PRESETS.contains(&value.trim()) {
+                        return Err(AppError::BadRequest(format!("Ungültiges Theme: {value}.")));
+                    }
+                    theme_changed = true;
+                }
+                "theme_primary" | "theme_accent" => {
+                    if !is_hex_color(value.trim()) {
+                        return Err(AppError::BadRequest(format!(
+                            "Ungültige Farbe für {key}: muss #rrggbb sein."
+                        )));
+                    }
+                    theme_changed = true;
+                }
+                _ => {}
             }
         } else {
             return Err(AppError::BadRequest(format!(
@@ -217,16 +291,30 @@ pub async fn patch_config(
 
     // Notify all clients that a publicly-readable config value changed so their stores
     // (e.g. the privacy note in My Account) refresh without a manual reload.
-    if privacy_note_changed {
+    if privacy_note_changed || theme_changed {
+        let mut keys: Vec<&str> = Vec::new();
+        if privacy_note_changed {
+            keys.push("privacy_note");
+        }
+        if theme_changed {
+            keys.push("theme");
+        }
         let _ = state.sse_tx.send(crate::state::SseEvent::new(
             "event-updated",
-            serde_json::json!({ "keys": ["privacy_note"] }).to_string(),
+            serde_json::json!({ "keys": keys }).to_string(),
         ));
     }
 
     Ok(StatusCode::NO_CONTENT)
 }
 
+/// A strict `#rrggbb` hex-colour check (6 hex digits, leading `#`). Deliberately not
+/// accepting shorthand/`#rgba` so the value is safe to drop straight into CSS.
+fn is_hex_color(s: &str) -> bool {
+    let bytes = s.as_bytes();
+    bytes.len() == 7 && bytes[0] == b'#' && bytes[1..].iter().all(|b| b.is_ascii_hexdigit())
+}
+
 pub async fn get_export_jobs(
     State(state): State<AppState>,
     RequireAdmin(_auth): RequireAdmin,
@@ -261,38 +349,108 @@ pub struct DownloadQuery {
 /// carry an `Authorization` header, so the client exchanges its Bearer token for
 /// an opaque ticket here, then hits `/export/zip?ticket=...`. Reuses the same
 /// single-use, 30s-TTL store as the SSE stream.
+#[derive(serde::Deserialize)]
+pub struct ExportTicketQuery {
+    /// Which archive the ticket is for — `zip` or `html`. Optional so an older client that
+    /// doesn't send it keeps working; it simply skips the pre-check it doesn't know to ask for.
+    #[serde(default)]
+    pub kind: Option<String>,
+}
+
 pub async fn export_ticket(
     State(state): State<AppState>,
+    axum::extract::Query(q): axum::extract::Query<ExportTicketQuery>,
     auth: crate::auth::middleware::AuthUser,
-) -> Json<serde_json::Value> {
+) -> Result<Json<serde_json::Value>, AppError> {
     // NOTE: intentionally NOT gated on `is_banned`. A banned user keeps *read* access
     // by design (USER_JOURNEYS §10.3, FEATURES: "Can still download the export once
     // released — Spec design choice"). The export is read-only, so it stays available
     // to them, consistent with the read-only-ban model.
-    let ticket = state.sse_tickets.issue(auth.token_hash);
-    Json(serde_json::json!({ "ticket": ticket }))
+
+    // The rate limit is enforced HERE rather than on the download itself, and that placement is
+    // the whole point: the download is an iframe navigation, so its response is invisible to the
+    // page. Limiting it there meant a guest over the limit tapped "Herunterladen", the ticket
+    // POST returned 200, the iframe silently received a 429, and absolutely nothing happened —
+    // forever, with no explanation, on the one screen that is the emotional payoff of the app.
+    // Minting is a normal `fetch`, so a 429 here reaches the user as a German message.
+    //
+    // Moving it does not weaken the limit: tickets are single-use with a 30s TTL and can only be
+    // obtained from this authenticated endpoint, so one mint is at most one download.
+    // Confirm the archive actually EXISTS before spending anything on it.
+    //
+    // `export_status` — which is what enables the Download button — reports `done` from
+    // `export_job`, while the download resolves through `export_current.file_path` plus a
+    // `Path::exists()`. Those are different sources of truth and can legitimately disagree: a
+    // row can say done while the file is gone, or an epoch bump can retire it between the page
+    // rendering and the guest tapping. When they disagreed the guest got the worst possible
+    // shape of failure — a green "Download gestartet" toast, a consumed single-use ticket, one
+    // of only three daily slots spent, and nothing in their Downloads folder, repeatable until
+    // the day's allowance was gone.
+    //
+    // Checking here, before `enforce_export_rate`, turns that into an honest error on a plain
+    // `fetch` that the existing `toastError` path already renders. This is NOT the HEAD probe
+    // ruled out elsewhere: it reads the same indexed row the download will read and touches no
+    // ticket, so it cannot consume anything.
+    if let Some(kind) = q.kind.as_deref() {
+        let export_type = match kind {
+            "zip" => "zip",
+            "html" => "html",
+            other => {
+                return Err(AppError::BadRequest(format!(
+                    "Unbekannter Export-Typ: {other}"
+                )));
+            }
+        };
+        let msg = if export_type == "zip" {
+            "Der ZIP-Export ist noch nicht verfügbar."
+        } else {
+            "Der HTML-Export ist noch nicht verfügbar."
+        };
+        resolve_export_file(&state, export_type, msg).await?;
+    }
+
+    enforce_export_rate(&state, auth.user_id).await?;
+
+    // `issue` returns None when the ticket store is at capacity. Unwrapping it into the JSON body
+    // serialized `{"ticket": null}` with a 200 — so `api.post` resolved happily, the page toasted
+    // success, the iframe navigated to `?ticket=null`, and one of the guest's three DAILY
+    // downloads had already been charged above. That is precisely the phantom-success failure
+    // this endpoint's pre-validation was added to eliminate, arriving through the other door.
+    // 503 + Retry-After, matching how `sse::issue_ticket` answers the identical condition.
+    let ticket = state
+        .sse_tickets
+        .issue(auth.token_hash, TicketKind::Download)
+        .ok_or_else(|| {
+            AppError::ServiceUnavailable(
+                "Server ist gerade ausgelastet. Bitte versuch es in einem Moment erneut.".into(),
+                Some(30),
+            )
+        })?;
+    Ok(Json(serde_json::json!({ "ticket": ticket })))
 }
 
 /// Validate a download ticket (single-use) and confirm its session still exists.
-async fn authenticate_download_ticket(state: &AppState, ticket: &str) -> Result<(), AppError> {
+/// Resolve a single-use download ticket to the user who minted it. The caller needs the
+/// id to key the export rate limit per-user (see `enforce_export_rate`).
+async fn authenticate_download_ticket(state: &AppState, ticket: &str) -> Result<Uuid, AppError> {
     let token_hash = state
         .sse_tickets
-        .consume(ticket)
+        .consume(ticket, TicketKind::Download)
         .ok_or_else(|| AppError::Unauthorized("Ticket ungültig oder abgelaufen.".into()))?;
-    crate::models::session::Session::find_by_token_hash(&state.pool, &token_hash)
+    let session = crate::models::session::Session::find_by_token_hash(&state.pool, &token_hash)
         .await
         .map_err(|e| AppError::Internal(e.into()))?
         .ok_or_else(|| AppError::Unauthorized("Sitzung nicht gefunden.".into()))?;
-    Ok(())
+    Ok(session.user_id)
 }
 
 pub async fn download_zip(
     State(state): State<AppState>,
     Query(q): Query<DownloadQuery>,
-    headers: HeaderMap,
 ) -> Result<axum::response::Response, AppError> {
+    // Ticket validation only — the rate limit was charged at mint time, where a 429 is visible
+    // to the page. Charging it again here would cost every download two slots.
     authenticate_download_ticket(&state, &q.ticket).await?;
-    enforce_export_rate(&state, &headers).await?;
 
     let path =
         resolve_export_file(&state, "zip", "Der ZIP-Export ist noch nicht verfügbar.").await?;
@@ -343,10 +501,9 @@ async fn resolve_export_file(
 pub async fn download_html(
     State(state): State<AppState>,
     Query(q): Query<DownloadQuery>,
-    headers: HeaderMap,
 ) -> Result<axum::response::Response, AppError> {
+    // See `download_zip`: the limit is charged at ticket mint, where the client can see it.
     authenticate_download_ticket(&state, &q.ticket).await?;
-    enforce_export_rate(&state, &headers).await?;
 
     let path =
         resolve_export_file(&state, "html", "Der HTML-Export ist noch nicht verfügbar.").await?;
@@ -403,8 +560,13 @@ pub async fn export_status(
     // worker superseded mid-run) is meaningless — surfacing its frozen `running`/77% would show a
     // progress bar that never moves for a keepsake nobody is building. It reads as "locked" (no
     // current job), which is exactly what it is.
-    let jobs: Vec<(String, String, i16)> = sqlx::query_as(
-        "SELECT j.type::text, j.status::text, j.progress_pct
+    // `error_message` is carried here, not just on the admin dashboard's job list. The host is the
+    // one who releases the keepsake and the one who owns the "Erneut versuchen" button, but this
+    // endpoint used to hand them a bare `failed` — so a fully actionable reason (notably the disk
+    // preflight's "needs X GB, Y GB free") was written to the row and then shown to nobody who
+    // could act on it. An admin-only diagnostic is not a diagnostic for the person on the spot.
+    let jobs: Vec<(String, String, i16, Option<String>)> = sqlx::query_as(
+        "SELECT j.type::text, j.status::text, j.progress_pct, j.error_message
            FROM export_job j
            JOIN event e ON e.id = j.event_id
           WHERE e.id = $1 AND j.epoch = e.export_epoch",
@@ -415,9 +577,21 @@ pub async fn export_status(
 
     let job_status = |type_name: &str| {
         jobs.iter()
-            .find(|(t, _, _)| t == type_name)
-            .map(|(_, status, pct)| serde_json::json!({ "status": status, "progress_pct": pct }))
-            .unwrap_or_else(|| serde_json::json!({ "status": "locked", "progress_pct": 0 }))
+            .find(|(t, _, _, _)| t == type_name)
+            .map(|(_, status, pct, err)| {
+                serde_json::json!({
+                    "status": status,
+                    "progress_pct": pct,
+                    // Only on a failure. A stale message left on a row that has since been re-armed
+                    // would otherwise show an error next to a running progress bar.
+                    "error_message": if status == "failed" { err.clone() } else { None },
+                })
+            })
+            .unwrap_or_else(|| {
+                serde_json::json!({
+                    "status": "locked", "progress_pct": 0, "error_message": null,
+                })
+            })
     };
 
     Ok(Json(serde_json::json!({
@@ -430,21 +604,29 @@ pub async fn export_status(
 /// Centralised guard for the export rate limit. Same pattern as upload/feed: master
 /// switch + per-endpoint switch + numeric value, all stored in `config` and read on
 /// each request.
-async fn enforce_export_rate(state: &AppState, headers: &HeaderMap) -> Result<(), AppError> {
+async fn enforce_export_rate(state: &AppState, user_id: Uuid) -> Result<(), AppError> {
     let rate_limits_on = config::get_bool(&state.config_cache, "rate_limits_enabled", true).await;
     let export_rate_on = config::get_bool(&state.config_cache, "export_rate_enabled", true).await;
     if !(rate_limits_on && export_rate_on) {
         return Ok(());
     }
-    let ip = client_ip(headers, "unknown");
     let limit = config::get_usize(&state.config_cache, "export_rate_per_day", 3).await;
-    if !state
-        .rate_limiter
-        .check(format!("export:{ip}"), limit, Duration::from_secs(86400))
-    {
+    // Keyed per-user. This was the worst of the IP-keyed limiters: 3 downloads per DAY
+    // shared across every guest behind the venue's public IP, so the fourth person to
+    // fetch their keepsake was locked out until the next day.
+    if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
+        format!("export:{user_id}"),
+        limit,
+        Duration::from_secs(86400),
+    ) {
+        // Names the real window. The generic "warte kurz" wording this used to share with the
+        // per-minute limiters is actively wrong here — the bucket is a DAY, so a guest told to
+        // wait a moment would keep tapping a button that cannot work again until tomorrow.
         return Err(AppError::TooManyRequests(
-            "Zu viele Anfragen. Bitte warte kurz und versuche es erneut.".into(),
-            None,
+            "Du hast das Tageslimit für Downloads erreicht. Versuch es später noch einmal — \
+             deine Galerie bleibt gespeichert."
+                .into(),
+            Some(retry_after_secs),
         ));
     }
     Ok(())

backend/src/handlers/feed.rs

@@ -2,7 +2,6 @@ use std::time::Duration;
 
 use axum::Json;
 use axum::extract::{Query, State};
-use axum::http::HeaderMap;
 use chrono::{DateTime, Utc};
 use serde::{Deserialize, Serialize};
 use uuid::Uuid;
@@ -10,14 +9,49 @@ use uuid::Uuid;
 use crate::auth::middleware::AuthUser;
 use crate::error::AppError;
 use crate::services::config;
-use crate::services::rate_limiter::client_ip;
 use crate::state::AppState;
 
 #[derive(Deserialize)]
 pub struct FeedQuery {
     pub cursor: Option<Uuid>,
     pub limit: Option<i64>,
+    /// Single tag (list view). Kept alongside `hashtags` so existing callers keep working.
     pub hashtag: Option<String>,
+    /// Comma-separated tags, combined with **OR** — the grid's chip semantics
+    /// (USER_JOURNEYS §8). Filtering moved server-side because the client could only ever
+    /// filter the pages it had already loaded: with page 1 = 20 items out of a 1000-photo
+    /// event, selecting a tag showed a handful of tiles and looked complete. Matching is
+    /// now the exact `hashtag` row in both views, so the grid and the list can no longer
+    /// disagree about which photos carry a tag (the client matched a caption SUBSTRING, so
+    /// `#tanz` also matched `#tanzflaeche`).
+    pub hashtags: Option<String>,
+    /// Exact uploader display name, combined with the tag group using **AND**.
+    pub uploader: Option<String>,
+}
+
+/// Merge the single-tag and CSV tag params into one normalised, de-duplicated list.
+///
+/// Normalisation mirrors `Hashtag::upsert` exactly (trim, drop a leading `#`, lowercase), so
+/// a chip built from a display string like `#Tanz` matches the stored `tanz` row. Returns
+/// `None` when no usable tag was supplied, which makes the SQL predicate a no-op — an empty
+/// list must mean "no tag filter", never "match nothing".
+fn normalize_tags(single: Option<&str>, csv: Option<&str>) -> Option<Vec<String>> {
+    let mut out: Vec<String> = Vec::new();
+    let mut push = |raw: &str| {
+        let t = raw.trim().trim_start_matches('#').to_lowercase();
+        if !t.is_empty() && !out.contains(&t) {
+            out.push(t);
+        }
+    };
+    if let Some(s) = single {
+        push(s);
+    }
+    if let Some(s) = csv {
+        for part in s.split(',') {
+            push(part);
+        }
+    }
+    if out.is_empty() { None } else { Some(out) }
 }
 
 #[derive(Serialize)]
@@ -27,6 +61,8 @@ pub struct FeedUpload {
     pub uploader_name: String,
     pub preview_url: Option<String>,
     pub thumbnail_url: Option<String>,
+    /// Big-screen (~2048px) variant for the diashow. Absent until the derivative exists.
+    pub display_url: Option<String>,
     pub mime_type: String,
     pub caption: Option<String>,
     pub like_count: i64,
@@ -48,6 +84,7 @@ struct FeedRow {
     uploader_name: String,
     preview_path: Option<String>,
     thumbnail_path: Option<String>,
+    display_path: Option<String>,
     mime_type: String,
     caption: Option<String>,
     like_count: i64,
@@ -58,26 +95,30 @@ struct FeedRow {
 pub async fn feed(
     State(state): State<AppState>,
     auth: AuthUser,
-    headers: HeaderMap,
     Query(q): Query<FeedQuery>,
 ) -> Result<Json<FeedResponse>, AppError> {
-    let ip = client_ip(&headers, "unknown");
     let rate_limits_on = config::get_bool(&state.config_cache, "rate_limits_enabled", true).await;
     let feed_rate_on = config::get_bool(&state.config_cache, "feed_rate_enabled", true).await;
     if rate_limits_on && feed_rate_on {
         let rate_limit = config::get_usize(&state.config_cache, "feed_rate_per_min", 60).await;
-        if !state
-            .rate_limiter
-            .check(format!("feed:{ip}"), rate_limit, Duration::from_secs(60))
-        {
+        // Keyed per-user, exactly like `feed_delta` below: at a venue every guest shares
+        // one public IP, so an IP key gave the whole party a single 60/min bucket and the
+        // fastest scroller starved everyone else.
+        if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
+            format!("feed:{}", auth.user_id),
+            rate_limit,
+            Duration::from_secs(60),
+        ) {
             return Err(AppError::TooManyRequests(
                 "Zu viele Anfragen. Bitte warte kurz und versuche es erneut.".into(),
-                None,
+                Some(retry_after_secs),
             ));
         }
     }
 
-    let limit = q.limit.unwrap_or(20).min(100);
+    // Clamped at BOTH ends: only the upper bound was enforced, so `?limit=-5` reached Postgres
+    // as `LIMIT -4` and answered a hand-written URL with a 500.
+    let limit = q.limit.unwrap_or(20).clamp(1, 100);
 
     // Resolve the cursor to a (created_at, id) position. The pair is compared as a
     // tuple so ties on created_at break on id — keyset pagination on created_at
@@ -90,43 +131,42 @@ pub async fn feed(
         None => (None, None),
     };
 
-    let rows = if let Some(hashtag) = &q.hashtag {
-        let tag = hashtag.trim().trim_start_matches('#').to_lowercase();
-        sqlx::query_as::<_, FeedRow>(
-            "SELECT v.id, v.user_id, v.uploader_name, v.preview_path, v.thumbnail_path,
-                    v.mime_type, v.caption, v.like_count, v.comment_count, v.created_at
-             FROM v_feed v
-             JOIN upload_hashtag uh ON uh.upload_id = v.id
-             JOIN hashtag h ON h.id = uh.hashtag_id AND h.tag = $1
-             WHERE v.event_id = $2
-               AND ($3::timestamptz IS NULL OR (v.created_at, v.id) < ($3, $4))
-             ORDER BY v.created_at DESC, v.id DESC
-             LIMIT $5",
-        )
-        .bind(&tag)
-        .bind(auth.event_id)
-        .bind(cursor_time)
-        .bind(cursor_id)
-        .bind(limit + 1)
-        .fetch_all(&state.pool)
-        .await?
-    } else {
-        sqlx::query_as::<_, FeedRow>(
-            "SELECT id, user_id, uploader_name, preview_path, thumbnail_path,
-                    mime_type, caption, like_count, comment_count, created_at
-             FROM v_feed
-             WHERE event_id = $1
-               AND ($2::timestamptz IS NULL OR (created_at, id) < ($2, $3))
-             ORDER BY created_at DESC, id DESC
-             LIMIT $4",
-        )
-        .bind(auth.event_id)
-        .bind(cursor_time)
-        .bind(cursor_id)
-        .bind(limit + 1)
-        .fetch_all(&state.pool)
-        .await?
-    };
+    // Tags from either param, normalised the same way `Hashtag::upsert` stores them
+    // (trimmed, leading `#` dropped, lowercased) so the comparison is exact.
+    let tags = normalize_tags(q.hashtag.as_deref(), q.hashtags.as_deref());
+    let uploader = q
+        .uploader
+        .as_deref()
+        .map(str::trim)
+        .filter(|s| !s.is_empty());
+
+    // ONE statement for every combination, rather than a branch per filter. `EXISTS` with
+    // `= ANY($4)` gives OR across the tag group without the row multiplication a JOIN would
+    // cause when a photo carries two selected tags; the uploader predicate ANDs on top. Both
+    // are no-ops when NULL, so the unfiltered feed takes the same path.
+    let rows = sqlx::query_as::<_, FeedRow>(
+        "SELECT v.id, v.user_id, v.uploader_name, v.preview_path, v.thumbnail_path,
+                v.display_path, v.mime_type, v.caption, v.like_count, v.comment_count,
+                v.created_at
+         FROM v_feed v
+         WHERE v.event_id = $1
+           AND ($2::timestamptz IS NULL OR (v.created_at, v.id) < ($2, $3))
+           AND ($4::text[] IS NULL OR EXISTS (
+                 SELECT 1 FROM upload_hashtag uh
+                 JOIN hashtag h ON h.id = uh.hashtag_id
+                 WHERE uh.upload_id = v.id AND h.tag = ANY($4)))
+           AND ($5::text IS NULL OR v.uploader_name = $5)
+         ORDER BY v.created_at DESC, v.id DESC
+         LIMIT $6",
+    )
+    .bind(auth.event_id)
+    .bind(cursor_time)
+    .bind(cursor_id)
+    .bind(tags.as_deref())
+    .bind(uploader)
+    .bind(limit + 1)
+    .fetch_all(&state.pool)
+    .await?;
 
     let has_more = rows.len() as i64 > limit;
     let rows: Vec<FeedRow> = rows.into_iter().take(limit as usize).collect();
@@ -154,6 +194,10 @@ pub async fn feed(
                 .thumbnail_path
                 .as_ref()
                 .map(|_| format!("/api/v1/upload/{}/thumbnail", r.id));
+            let display_url = r
+                .display_path
+                .as_ref()
+                .map(|_| format!("/api/v1/upload/{}/display", r.id));
             FeedUpload {
                 liked_by_me: liked_set.contains(&r.id),
                 id: r.id,
@@ -161,6 +205,7 @@ pub async fn feed(
                 uploader_name: r.uploader_name,
                 preview_url,
                 thumbnail_url,
+                display_url,
                 mime_type: r.mime_type,
                 caption: r.caption,
                 like_count: r.like_count,
@@ -216,14 +261,14 @@ pub async fn feed_delta(
     let feed_rate_on = config::get_bool(&state.config_cache, "feed_rate_enabled", true).await;
     if rate_limits_on && feed_rate_on {
         let rate_limit = config::get_usize(&state.config_cache, "feed_rate_per_min", 60).await;
-        if !state.rate_limiter.check(
+        if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
             format!("feed_delta:{}", auth.user_id),
             rate_limit,
             Duration::from_secs(60),
         ) {
             return Err(AppError::TooManyRequests(
                 "Zu viele Anfragen. Bitte warte kurz und versuche es erneut.".into(),
-                None,
+                Some(retry_after_secs),
             ));
         }
     }
@@ -247,7 +292,7 @@ pub async fn feed_delta(
     // response's `server_time`, so this doesn't re-fetch on every subsequent delta.
     let rows = sqlx::query_as::<_, FeedRow>(
         "SELECT id, user_id, uploader_name, preview_path, thumbnail_path,
-                mime_type, caption, like_count, comment_count, created_at
+                display_path, mime_type, caption, like_count, comment_count, created_at
          FROM v_feed
          WHERE event_id = $1 AND created_at >= $2
          ORDER BY created_at DESC, id DESC
@@ -304,6 +349,10 @@ pub async fn feed_delta(
                 .thumbnail_path
                 .as_ref()
                 .map(|_| format!("/api/v1/upload/{}/thumbnail", r.id)),
+            display_url: r
+                .display_path
+                .as_ref()
+                .map(|_| format!("/api/v1/upload/{}/display", r.id)),
             mime_type: r.mime_type,
             caption: r.caption,
             like_count: r.like_count,
@@ -344,6 +393,32 @@ pub async fn hashtags(
     ))
 }
 
+/// Every uploader who has at least one visible upload, for the grid's "Nutzer suchen" picker.
+///
+/// The picker used to derive names from the uploads currently in memory — page 1, 20 items —
+/// so typing a guest's name found nothing whenever their photos happened to sit below the
+/// fold, which reads as "search is broken". This is the authoritative list.
+///
+/// Reads `v_feed`, so it inherits exactly the feed's visibility rules: soft-deleted uploads,
+/// banned uploaders and hidden uploaders are all excluded, and a guest who has not uploaded
+/// anything never appears. Uncapped on purpose — one short string per uploader, bounded by
+/// the guest count, and truncating it would reintroduce the very bug this replaces.
+/// Deliberately NOT the host-only `/host/users` route: that one lists every joined guest and
+/// exposes moderation state.
+pub async fn uploaders(
+    State(state): State<AppState>,
+    auth: AuthUser,
+) -> Result<Json<Vec<String>>, AppError> {
+    let rows: Vec<(String,)> = sqlx::query_as(
+        "SELECT DISTINCT uploader_name FROM v_feed WHERE event_id = $1 ORDER BY uploader_name",
+    )
+    .bind(auth.event_id)
+    .fetch_all(&state.pool)
+    .await?;
+
+    Ok(Json(rows.into_iter().map(|(name,)| name).collect()))
+}
+
 /// Resolve a cursor id to its `(created_at, id)` position. Both are needed:
 /// `created_at` alone isn't unique, so pagination must break ties on `id` to
 /// avoid silently dropping rows that share a timestamp across a page boundary.
@@ -375,3 +450,41 @@ async fn get_liked_set(
 
     rows.into_iter().map(|r| r.0).collect()
 }
+
+#[cfg(test)]
+mod tests {
+    use super::normalize_tags;
+
+    /// The chips carry display strings (`#Tanz`), the `hashtag` table stores `tanz`. If these
+    /// two drift the filter silently returns nothing, which is indistinguishable from "no
+    /// photos have this tag" — so pin the normalisation to `Hashtag::upsert`'s rule.
+    #[test]
+    fn tags_are_normalised_like_upsert_stores_them() {
+        assert_eq!(
+            normalize_tags(Some("#Tanz"), None),
+            Some(vec!["tanz".to_string()])
+        );
+        assert_eq!(
+            normalize_tags(None, Some(" #Buffet , reden ")),
+            Some(vec!["buffet".to_string(), "reden".to_string()])
+        );
+    }
+
+    /// An empty list must mean "no filter", never "match nothing" — returning `Some(vec![])`
+    /// would make `= ANY('{}')` false for every row and blank the feed.
+    #[test]
+    fn blank_input_disables_the_filter() {
+        assert_eq!(normalize_tags(None, None), None);
+        assert_eq!(normalize_tags(Some("   "), Some(" , ,#")), None);
+    }
+
+    /// Both params feed one list, de-duplicated: the list view sends `hashtag`, the grid sends
+    /// `hashtags`, and carrying a filter across views can legitimately set both to the same tag.
+    #[test]
+    fn single_and_csv_merge_without_duplicates() {
+        assert_eq!(
+            normalize_tags(Some("tanz"), Some("tanz,buffet")),
+            Some(vec!["tanz".to_string(), "buffet".to_string()])
+        );
+    }
+}

backend/src/handlers/host.rs

@@ -35,6 +35,46 @@ pub struct EventStatus {
     pub is_active: bool,
     pub uploads_locked: bool,
     pub export_released: bool,
+    /// Free space on the volume the keepsake is written to. `None` when the mount can't be
+    /// resolved — the UI hides the widget rather than rendering a confident zero.
+    pub disk_free_bytes: Option<u64>,
+    /// What a full keepsake build would need right now (both halves).
+    pub keepsake_required_bytes: u64,
+    /// Whether the host should be warned. See [`disk_is_low`].
+    pub disk_low: bool,
+}
+
+
+/// Is free space low enough that the host needs to know?
+///
+/// Two triggers, because a fixed threshold answers the wrong question. `postgres_data`,
+/// `media_data` and `exports_data` are all Docker named volumes on one filesystem, so a full disk
+/// does not degrade one subsystem — it stops Postgres writing and takes the event down. That is
+/// what the absolute floor is for.
+///
+/// The second trigger is the one that actually earns its place: the keepsake needs room for two
+/// gallery-sized archives, and the only moment a host can do anything about that is BEFORE they
+/// release. Warning at "you could not build the keepsake right now" turns a post-event dead end
+/// into a decision someone can still make.
+///
+/// IT MUST FIRE BEFORE THE UPLOAD GATE CLOSES, and that is why the reserve and the margin are
+/// here. The gate in `handlers::upload` refuses at `free < keepsake_required + DISK_RESERVE_BYTES`;
+/// warning at `free < keepsake_required` alone meant the two differed by the whole reserve, so
+/// the wall was always hit FIRST. Every guest would be blocked from uploading while this
+/// dashboard showed a comfortable disk and no banner at all — on the shipped 40 GB box, uploads
+/// stopping with ~27 GB free and nothing on screen to explain it, with no operator present.
+///
+/// The 25% margin makes it a warning rather than an obituary: the host sees it while there is
+/// still room to act (delete a few large videos, which refunds immediately and reopens the gate).
+fn disk_is_low(free: u64, keepsake_required: u64) -> bool {
+    let gate_closes_at = keepsake_required.saturating_add(crate::handlers::upload::DISK_RESERVE_BYTES as u64);
+    let warn_at = gate_closes_at.saturating_add(gate_closes_at / 4);
+    // No separate absolute-floor clause. There used to be `free < LOW_DISK_FLOOR_BYTES ||` here,
+    // and it was unreachable: `gate_closes_at` is at least DISK_RESERVE_BYTES, so `warn_at` is at
+    // least 1.25x it (12.5 GB) — always above the 10 GB floor. Two tests were named after that
+    // clause and neither could fail if it were deleted. Keeping dead code that tests claim to
+    // cover is worse than not having it.
+    free < warn_at
 }
 
 /// Count non-banned hosts/admins in the event OTHER than `excluding` — the operators
@@ -72,11 +112,29 @@ pub async fn get_event_status(
         .await?
         .ok_or_else(|| AppError::NotFound("Event nicht gefunden.".into()))?;
 
+    // Measured on the EXPORT volume, not the media one: that is where the cliff is, and it is a
+    // distinct mount point even when both are backed by the same filesystem. The cached reading is
+    // right here — this is advisory, polled on every dashboard load, and a 15s-stale number costs
+    // nothing (unlike the export preflight, which reads uncached because it is about to write).
+    let free = state
+        .disk_cache
+        .snapshot(&state.config.export_path)
+        .map(|d| d.free);
+    let keepsake_required_bytes =
+        crate::services::export::keepsake_space_required(&state.pool, event.id)
+            .await
+            .unwrap_or(0);
+
     Ok(Json(EventStatus {
         name: event.name,
         is_active: event.is_active,
         uploads_locked: event.uploads_locked_at.is_some(),
         export_released: event.export_released_at.is_some(),
+        disk_free_bytes: free,
+        keepsake_required_bytes,
+        // Unknown free space is NOT low. Fails open, exactly as the upload quota and the export
+        // preflight do: a scary banner on an unreadable mount would train the host to ignore it.
+        disk_low: free.is_some_and(|f| disk_is_low(f, keepsake_required_bytes)),
     }))
 }
 
@@ -302,6 +360,7 @@ pub async fn rebuild_export(
         r.event_id,
         r.event_name,
         r.epoch,
+        state.config.comments_enabled,
         std::time::Duration::ZERO,
         state.pool.clone(),
         state.config.media_path.clone(),
@@ -432,7 +491,7 @@ pub async fn reset_user_pin(
     }
 
     let pin: String = format!("{:04}", rand::rng().random_range(0..10000u32));
-    let pin_hash = bcrypt::hash(&pin, 12).map_err(|e| AppError::Internal(anyhow::anyhow!(e)))?;
+    let pin_hash = crate::auth::handlers::hash_password(pin.clone(), 12).await?;
 
     sqlx::query(
         "UPDATE \"user\"
@@ -549,6 +608,7 @@ pub fn start_regen(state: &AppState, regen: crate::services::export::PendingRege
         regen.event_id,
         regen.event_name,
         regen.epoch,
+        state.config.comments_enabled,
         // Debounced: a takedown pass is a burst, and each request retires the last generation. The
         // delay lets superseded workers fail their claim and do zero work instead of each building
         // a full archive. See export::REGEN_DEBOUNCE.
@@ -753,6 +813,7 @@ pub async fn release_gallery(
         event_id,
         event_name,
         epoch,
+        state.config.comments_enabled,
         std::time::Duration::ZERO,
         state.pool.clone(),
         state.config.media_path.clone(),
@@ -762,3 +823,99 @@ pub async fn release_gallery(
 
     Ok(StatusCode::NO_CONTENT)
 }
+
+#[cfg(test)]
+mod tests {
+    use super::disk_is_low;
+    use crate::handlers::upload::DISK_RESERVE_BYTES;
+    use crate::services::export::required_free_bytes;
+
+    const GB: u64 = 1_000_000_000;
+
+    #[test]
+    fn a_healthy_disk_with_room_for_the_keepsake_is_not_low() {
+        // Room for the keepsake AND the reserve the upload gate holds back, with margin.
+        assert!(!disk_is_low(60 * GB, 25 * GB));
+    }
+
+    /// Renamed from `the_absolute_floor_fires_...`: there is no separate floor clause any more
+    /// (see `disk_is_low`). What still has to hold is the behaviour the floor was there FOR — a
+    /// nearly-empty disk is low even when the gallery is small enough that the keepsake term
+    /// alone would clear it, because all three volumes share one filesystem and Postgres needs
+    /// room to write.
+    #[test]
+    fn a_nearly_empty_disk_is_low_even_when_the_gallery_is_tiny() {
+        // All three volumes share one filesystem, so running out doesn't degrade one subsystem —
+        // Postgres stops being able to write and the event goes down. A 1 GB gallery would clear
+        // the keepsake test comfortably; the floor is what catches this.
+        assert!(disk_is_low(5 * GB, GB));
+        assert!(disk_is_low(9 * GB, 0));
+        assert!(!disk_is_low(20 * GB, 0), "a roomy empty disk is not low");
+    }
+
+    /// THE PROPERTY THIS EXISTS FOR: the host must be warned BEFORE guests are blocked.
+    ///
+    /// `handlers::upload` refuses at `free < keepsake_required + DISK_RESERVE_BYTES`. If the
+    /// banner fires only at or below that, the host's first signal is 100 guests being unable
+    /// to upload while the dashboard shows a comfortable disk — with nobody on site to ask.
+    #[test]
+    fn the_banner_always_fires_before_the_upload_gate_closes() {
+        // Asserting `disk_is_low(gate_closes_at, required)` is what this used to do, and it was a
+        // tautology: `disk_is_low` recomputes the same `gate_closes_at` internally and compares
+        // against `gate + gate/4`, so the assertion reduced to `G < G + G/4` — true for every G,
+        // for any margin, even a margin of zero. It could not detect the banner being moved to
+        // exactly the gate, which is the regression it is named for.
+        //
+        // So pin the GAP instead: find the free-space level at which the banner starts, and
+        // require it to be strictly above the level at which the gate closes, by a usable amount.
+        for media_gb in [0u64, 1, 4, 8, 16, 32] {
+            let required = required_free_bytes(media_gb * GB, 2);
+            let gate_closes_at = required + DISK_RESERVE_BYTES as u64;
+
+            // Just above the gate: guests can still upload, and the host must already be warned.
+            assert!(
+                disk_is_low(gate_closes_at + 1, required),
+                "at media={media_gb}GB the banner is not yet showing while the gate still allows uploads"
+            );
+
+            // The warning must lead by a margin the host can act inside, not by one byte.
+            let mut warn_starts_at = gate_closes_at;
+            while disk_is_low(warn_starts_at, required) {
+                warn_starts_at += GB / 10;
+            }
+            assert!(
+                warn_starts_at >= gate_closes_at + gate_closes_at / 5,
+                "at media={media_gb}GB the banner leads the gate by only                  {} bytes",
+                warn_starts_at - gate_closes_at
+            );
+        }
+    }
+
+    #[test]
+    fn plenty_of_space_is_still_low_when_the_keepsake_would_not_fit() {
+        // THE case the fixed threshold misses, and the one that matters: 30 GB free is nowhere near
+        // any floor, but a 30 GB gallery needs room for TWO archives. The host can act on this
+        // before releasing; after releasing, they cannot.
+        assert!(disk_is_low(30 * GB, 66 * GB));
+    }
+
+    #[test]
+    fn the_keepsake_trigger_is_exact_at_the_boundary() {
+        // The boundary is the UPLOAD GATE's threshold plus a 25% margin, not the bare keepsake
+        // size — see `disk_is_low`. Warning at the bare size fired only after the gate had
+        // already blocked every guest.
+        let required = 20 * GB;
+        let gate = required + DISK_RESERVE_BYTES as u64;
+        let warn_at = gate + gate / 4;
+        assert!(!disk_is_low(warn_at, required), "exactly enough is enough");
+        assert!(disk_is_low(warn_at - 1, required));
+    }
+
+    #[test]
+    fn an_empty_gallery_still_reserves_room_for_postgres() {
+        // With no gallery the keepsake term is 0, so the warn threshold collapses to
+        // 1.25 x DISK_RESERVE_BYTES (12.5 GB), which dominates the 10 GB absolute floor.
+        assert!(!disk_is_low(13 * GB, 0));
+        assert!(disk_is_low(9 * GB, 0));
+    }
+}

backend/src/handlers/me.rs

@@ -14,7 +14,7 @@ use serde::Serialize;
 use crate::auth::middleware::AuthUser;
 use crate::error::AppError;
 use crate::handlers::upload::compute_storage_quota;
-use crate::models::user::User;
+use crate::models::user::{User, UserRole};
 use crate::services::config;
 use crate::state::AppState;
 
@@ -37,12 +37,26 @@ pub async fn get_quota(
 
     let estimate = compute_storage_quota(&state).await;
 
+    // Raw server telemetry (free disk, concurrent uploader count) is staff-only — it
+    // must never reach a guest, even though the guest upload UI no longer renders it.
+    // A guest still gets their own `used`/`limit` so enforcement stays transparent to
+    // the code paths that consume it; only the server-wide fields are zeroed.
+    let is_staff = matches!(auth.role, UserRole::Host | UserRole::Admin);
+
     Ok(Json(QuotaDto {
         enabled: estimate.limit_bytes.is_some(),
         used_bytes: user.total_upload_bytes,
         limit_bytes: estimate.limit_bytes,
-        active_uploaders: estimate.active_uploaders,
-        free_disk_bytes: estimate.free_disk_bytes,
+        active_uploaders: if is_staff {
+            estimate.active_uploaders
+        } else {
+            0
+        },
+        free_disk_bytes: if is_staff {
+            estimate.free_disk_bytes
+        } else {
+            0
+        },
     }))
 }
 
@@ -61,6 +75,14 @@ pub struct MeContextDto {
     /// The gallery has been released and the export snapshotted — uploads are permanently
     /// closed for this run (release ⇒ lock, and reopening regenerates).
     pub gallery_released: bool,
+    /// This guest is banned: a deliberately READ-ONLY ban (see `handlers/host.rs`) — they keep
+    /// the feed and the keepsake, but every write is refused.
+    ///
+    /// Exposed so the UI can SAY so. Without it the client had no idea, so the upload button,
+    /// the like button and "Löschen" all rendered enabled and returned 403 "Du bist gesperrt."
+    /// on every tap — a guest tapping upload repeatedly with nobody to ask. The lock case
+    /// (`uploads_locked`) has always been surfaced for exactly this reason; a ban was not.
+    pub is_banned: bool,
 }
 
 pub async fn get_context(
@@ -96,5 +118,6 @@ pub async fn get_context(
         storage_quota_enabled,
         uploads_locked,
         gallery_released,
+        is_banned: user.is_banned,
     }))
 }

backend/src/handlers/public.rs

@@ -4,21 +4,41 @@ use axum::Json;
 use axum::extract::State;
 use serde::Serialize;
 
+use crate::services::config;
 use crate::state::AppState;
 
 #[derive(Serialize)]
 pub struct PublicEventDto {
     pub name: String,
     pub slug: String,
+    /// Whether the comment feature is on (env `COMMENTS_ENABLED`). The frontend hides
+    /// the whole comment UI when false; exposed here so even the pre-auth shell knows.
+    pub comments_enabled: bool,
+    /// Active colour theme. `preset` is an id the frontend maps to a palette (or
+    /// "custom"); `primary`/`accent` are the `#rrggbb` seeds the ramps derive from.
+    /// Resolved as DB-config override → env default. Public so the theme applies on
+    /// the very first (pre-auth) paint without a flash.
+    pub theme_preset: String,
+    pub theme_primary: String,
+    pub theme_accent: String,
 }
 
-/// Public event identity, used by the pre-auth join/recover screens so a guest can
-/// see *which* event they're joining. Only the display name and slug are exposed —
-/// nothing user-scoped — so this is safe without a token. Served straight from the
-/// instance config (no DB round-trip needed).
+/// Public event identity + presentation config, used by the pre-auth join/recover
+/// screens (which event am I joining, what does it look like). Only non-user-scoped
+/// fields are exposed, so this is safe without a token. Identity comes straight from
+/// instance config; the theme is resolved from the runtime `config` table (admin UI)
+/// falling back to the env-seeded default.
 pub async fn get_public_event(State(state): State<AppState>) -> Json<PublicEventDto> {
+    let cache = &state.config_cache;
     Json(PublicEventDto {
         name: state.config.event_name.clone(),
         slug: state.config.event_slug.clone(),
+        comments_enabled: state.config.comments_enabled,
+        theme_preset: config::get_str(cache, "theme_preset", &state.config.default_theme_preset)
+            .await,
+        theme_primary: config::get_str(cache, "theme_primary", &state.config.default_theme_primary)
+            .await,
+        theme_accent: config::get_str(cache, "theme_accent", &state.config.default_theme_accent)
+            .await,
     })
 }

backend/src/handlers/social.rs

@@ -10,8 +10,40 @@ use crate::error::AppError;
 use crate::models::comment::{Comment, CommentDto};
 use crate::models::hashtag::{self, Hashtag};
 use crate::models::upload::Upload;
+use crate::services::config;
 use crate::state::AppState;
 
+/// Throttle a social write. Keyed PER USER, like the feed and upload limits and for the same
+/// reason: at a venue every guest sits behind one NAT, so an IP key hands the whole party a
+/// single bucket and the most active guest starves everyone else.
+///
+/// These were the only mutating endpoints in the app with no limit at all — the coverage was
+/// asymmetric, not deliberately open. The ceiling is set well above anything a real guest
+/// produces; this bounds a script, not an enthusiastic double-tapper.
+async fn check_social_rate(state: &AppState, user_id: Uuid) -> Result<(), AppError> {
+    let rate_limits_on = config::get_bool(&state.config_cache, "rate_limits_enabled", true).await;
+    let social_rate_on = config::get_bool(&state.config_cache, "social_rate_enabled", true).await;
+    if !(rate_limits_on && social_rate_on) {
+        return Ok(());
+    }
+    let rate_limit = config::get_usize(&state.config_cache, "social_rate_per_min", 120).await;
+    // ONE bucket across likes, comments and comment deletions. Separate buckets would let a
+    // caller triple the aggregate write rate just by alternating between them.
+    state
+        .rate_limiter
+        .check_with_retry(
+            format!("social:{user_id}"),
+            rate_limit,
+            std::time::Duration::from_secs(60),
+        )
+        .map_err(|retry_after_secs| {
+            AppError::TooManyRequests(
+                "Zu viele Aktionen. Bitte warte kurz und versuche es erneut.".into(),
+                Some(retry_after_secs),
+            )
+        })
+}
+
 #[derive(Serialize)]
 pub struct LikeResponse {
     /// The caller's like state *after* this toggle. The client sets `liked_by_me` from
@@ -35,6 +67,7 @@ pub async fn toggle_like(
     if user.is_banned {
         return Err(AppError::Forbidden("Du bist gesperrt.".into()));
     }
+    check_social_rate(&state, auth.user_id).await?;
 
     // Event-scope: the upload must belong to the caller's event (404 otherwise),
     // matching the host handlers' find_by_id_and_event pattern.
@@ -129,12 +162,19 @@ pub async fn add_comment(
     Path(upload_id): Path<Uuid>,
     Json(body): Json<AddCommentRequest>,
 ) -> Result<(StatusCode, Json<CommentDto>), AppError> {
+    // Comments can be disabled instance-wide (env COMMENTS_ENABLED). The frontend hides
+    // the UI, but gate the API too so a stale client or direct call can't slip one in.
+    if !state.config.comments_enabled {
+        return Err(AppError::Forbidden("Kommentare sind deaktiviert.".into()));
+    }
+
     let user = crate::models::user::User::find_by_id(&state.pool, auth.user_id)
         .await?
         .ok_or_else(|| AppError::NotFound("Benutzer nicht gefunden.".into()))?;
     if user.is_banned {
         return Err(AppError::Forbidden("Du bist gesperrt.".into()));
     }
+    check_social_rate(&state, auth.user_id).await?;
 
     // Event-scope: only comment on an upload that belongs to the caller's event.
     Upload::find_by_id_and_event(&state.pool, upload_id, auth.event_id)
@@ -155,7 +195,14 @@ pub async fn add_comment(
 
     // Insert the comment and link its hashtags atomically, so a crash mid-loop
     // can't leave a committed comment with only some of its tags indexed.
-    let tags = hashtag::extract_hashtags(text);
+    let mut tags = hashtag::extract_hashtags(text);
+    // Deterministic lock order, matching the upload path. `Hashtag::upsert` takes row locks,
+    // so two transactions touching the same two tags in OPPOSITE order deadlock — Postgres
+    // aborts one after ~1s and that guest's comment 500s. `extract_hashtags` returns them in
+    // text order, which is exactly the unordered case. Sort on the NORMALISED form, because
+    // that is the key `upsert` locks on.
+    tags.sort_by_key(|t| t.trim().trim_start_matches('#').to_lowercase());
+    tags.dedup_by_key(|t| t.trim().trim_start_matches('#').to_lowercase());
     let mut tx = state.pool.begin().await?;
     let comment = Comment::create(&mut *tx, upload_id, auth.user_id, text).await?;
     for tag in &tags {
@@ -210,6 +257,7 @@ pub async fn delete_comment(
     if auth.is_banned {
         return Err(AppError::Forbidden("Du bist gesperrt.".into()));
     }
+    check_social_rate(&state, auth.user_id).await?;
     let comment = Comment::find_by_id(&state.pool, comment_id)
         .await?
         .ok_or_else(|| AppError::NotFound("Kommentar nicht gefunden.".into()))?;

backend/src/handlers/sse.rs

@@ -13,6 +13,7 @@ use tokio_stream::wrappers::errors::BroadcastStreamRecvError;
 use crate::auth::middleware::AuthUser;
 use crate::error::AppError;
 use crate::models::session::Session;
+use crate::services::sse_tickets::TicketKind;
 use crate::state::AppState;
 
 #[derive(Deserialize)]
@@ -38,7 +39,26 @@ pub async fn issue_ticket(
     State(state): State<AppState>,
     auth: AuthUser,
 ) -> Result<Json<StreamTicketResponse>, AppError> {
-    let ticket = state.sse_tickets.issue(auth.token_hash);
+    // The endpoint had no rate limit at all. Authentication is not a bound here: one valid
+    // session could loop it freely. 60/min is far above a real client (one ticket per SSE
+    // (re)connect, and reconnects are backed off) while capping a loop.
+    if let Err(retry_after_secs) = state.rate_limiter.check_with_retry(
+        format!("sse_ticket:{}", auth.user_id),
+        60,
+        Duration::from_secs(60),
+    ) {
+        return Err(AppError::TooManyRequests(
+            "Zu viele Verbindungsversuche. Bitte warte kurz.".into(),
+            Some(retry_after_secs),
+        ));
+    }
+
+    let ticket = state.sse_tickets.issue(auth.token_hash, TicketKind::Sse).ok_or_else(|| {
+        AppError::ServiceUnavailable(
+            "Server ist gerade ausgelastet. Live-Updates folgen in Kürze.".into(),
+            Some(30),
+        )
+    })?;
     let server_time = sqlx::query_scalar("SELECT NOW()")
         .fetch_one(&state.pool)
         .await?;
@@ -56,7 +76,7 @@ pub async fn stream(
 ) -> Result<Sse<impl Stream<Item = Result<Event, Infallible>>>, AppError> {
     let token_hash = state
         .sse_tickets
-        .consume(&q.ticket)
+        .consume(&q.ticket, TicketKind::Sse)
         .ok_or_else(|| AppError::Unauthorized("Ticket ungültig oder abgelaufen.".into()))?;
 
     // NOTE: this authenticates via ticket→session, not the `AuthUser` extractor. The

backend/src/handlers/test_admin.rs

@@ -13,9 +13,10 @@ use crate::auth::middleware::RequireAdmin;
 use crate::error::AppError;
 use crate::state::AppState;
 
-/// Truncates every event-scoped table, wipes media on disk, and reseeds the
-/// `config` table from migration defaults. Requires an admin JWT — even with
-/// `EVENTSNAP_TEST_MODE=1` it cannot be hit anonymously.
+/// Truncates every event-scoped table, wipes media on disk, and reseeds the `config`
+/// table: numeric values from the migration defaults, but every feature toggle forced
+/// OFF (production seeds them ON — see the note at the reseed below). Requires an admin
+/// JWT — even with `EVENTSNAP_TEST_MODE=1` it cannot be hit anonymously.
 pub async fn truncate_all(
     State(state): State<AppState>,
     RequireAdmin(_auth): RequireAdmin,
@@ -40,15 +41,29 @@ pub async fn truncate_all(
     .execute(&state.pool)
     .await?;
 
-    // Reseed config — mirrors migrations 005 and 009. Kept in sync by hand
-    // because pulling SQL out of the migration files at runtime is fragile.
+    // Reseed config. The NUMERIC values mirror migrations 005/015/016/017/019; the BOOLEAN
+    // toggles deliberately do NOT — migration 009 seeds every one of them `true`
+    // (production), and this forces them `false` so the suite isn't fighting rate limits
+    // and quotas it isn't testing.
+    //
+    // Be aware of what that costs: this runs as an auto-fixture before EVERY test, so no
+    // test starts from production's config unless it explicitly turns a toggle back on
+    // (02-upload/rate-limit, 07-adversarial/ddos, 01-auth/rate-limit-nat, …). That blind
+    // spot is exactly why an entire class of per-IP limiter bugs went unnoticed: the
+    // limiters were simply off. When adding a limiter or quota, add a spec that enables it.
+    //
+    // Kept in sync by hand because pulling SQL out of the migration files at runtime is
+    // fragile — if you add a config key in a migration, add it here too.
     sqlx::query(
         r#"INSERT INTO config (key, value) VALUES
             ('max_image_size_mb',          '20'),
             ('max_video_size_mb',          '500'),
-            ('upload_rate_per_hour',       '10'),
+            ('upload_rate_per_hour',       '100'),
             ('feed_rate_per_min',          '60'),
             ('export_rate_per_day',        '3'),
+            ('join_ip_rate_per_min',       '60'),
+            ('recover_ip_rate_per_min',    '30'),
+            ('social_rate_per_min',        '120'),
             ('quota_tolerance',            '0.75'),
             ('estimated_guest_count',      '100'),
             ('compression_concurrency',    '2'),
@@ -57,6 +72,8 @@ pub async fn truncate_all(
             ('feed_rate_enabled',          'false'),
             ('export_rate_enabled',        'false'),
             ('join_rate_enabled',          'false'),
+            ('social_rate_enabled',        'false'),
+            ('admin_login_rate_enabled',   'false'),
             ('quota_enabled',              'false'),
             ('storage_quota_enabled',      'false'),
             ('upload_count_quota_enabled', 'false'),
@@ -94,6 +111,12 @@ pub async fn truncate_all(
     // steers the per-user limit off `free_disk_bytes`), i.e. two holes were masking each other.
     state.disk_cache.invalidate();
 
+    // `media_total` caches SUM(user.total_upload_bytes) for the upload gate's keepsake-headroom
+    // check. TRUNCATE has just zeroed every one of those rows, so a surviving reading would make
+    // the next test's first upload measure its headroom against the previous test's gallery —
+    // and that gate REFUSES uploads, so the failure would look like a spurious quota rejection.
+    state.media_total.invalidate();
+
     // `sse_tickets` maps a ticket to a session token hash. TRUNCATE deletes the sessions, so every
     // surviving ticket is a dangling reference to a user that no longer exists.
     state.sse_tickets.clear();

backend/src/main.rs

@@ -2,7 +2,6 @@ use anyhow::Result;
 use axum::Router;
 use axum::extract::DefaultBodyLimit;
 use axum::routing::{delete, get, patch, post};
-use tower_http::services::ServeDir;
 use tower_http::trace::TraceLayer;
 use tracing_subscriber::{layer::SubscriberExt, util::SubscriberInitExt};
 
@@ -28,8 +27,15 @@ async fn main() -> Result<()> {
 
     tracing_subscriber::registry()
         .with(
+            // `info`, not `debug`. A stock deploy sets RUST_LOG nowhere (it is absent from
+            // .env.example and was absent from docker-compose.yml), so this fallback IS the
+            // production level — and at `debug` the TraceLayer below emits a line per request
+            // AND per response, into a log file that had no rotation. `tower_http=warn`
+            // rather than `info` states the intent: those spans are diagnostics, not an
+            // access log, and a future `DefaultOnResponse::new().level(Level::INFO)` should
+            // not silently re-enable them.
             tracing_subscriber::EnvFilter::try_from_default_env()
-                .unwrap_or_else(|_| "eventsnap_backend=debug,tower_http=debug".into()),
+                .unwrap_or_else(|_| "eventsnap_backend=info,tower_http=warn".into()),
         )
         .with(tracing_subscriber::fmt::layer())
         .init();
@@ -45,6 +51,19 @@ async fn main() -> Result<()> {
 
     let state = AppState::new(pool.clone(), config.clone());
 
+    // Regenerate image derivatives an older pipeline produced: the big-screen display for
+    // uploads processed before it existed (v0.17.x), and anything predating the current
+    // DERIVATIVES_REV (rev 1 applies the EXIF orientation, without which every portrait
+    // phone photo is stored sideways). Fire-and-forget behind the compression semaphore;
+    // originals are never touched, so a failure just retries on the next start.
+    state.compression.backfill_stale_derivatives().await;
+
+    // Re-extract poster frames for videos a restart interrupted. `startup_recovery` above
+    // marks their compression `failed` but nothing re-enqueued them, so `thumbnail_path`
+    // stayed NULL for the rest of the event. Shares the attempt budget with the image
+    // backfill, so a clip that genuinely yields no frame stops being retried.
+    state.compression.backfill_video_posters().await;
+
     // Re-spawn exports for events that were released but whose keepsake never finished
     // (crash mid-export). Needs the media/export paths + SSE sender, so it runs here
     // rather than inside `startup_recovery`. Fire-and-forget: the workers run in the
@@ -53,6 +72,7 @@ async fn main() -> Result<()> {
         pool.clone(),
         config.media_path.clone(),
         config.export_path.clone(),
+        config.comments_enabled,
         state.sse_tx.clone(),
     )
     .await;
@@ -63,6 +83,7 @@ async fn main() -> Result<()> {
         pool,
         state.rate_limiter.clone(),
         state.sse_tickets.clone(),
+        config.media_path.clone(),
     );
 
     // Ensure media directories exist
@@ -106,6 +127,10 @@ async fn main() -> Result<()> {
             "/api/v1/upload/{id}/preview",
             get(handlers::upload::get_preview),
         )
+        .route(
+            "/api/v1/upload/{id}/display",
+            get(handlers::upload::get_display),
+        )
         .route(
             "/api/v1/upload/{id}/thumbnail",
             get(handlers::upload::get_thumbnail),
@@ -117,6 +142,7 @@ async fn main() -> Result<()> {
         .route("/api/v1/feed", get(handlers::feed::feed))
         .route("/api/v1/feed/delta", get(handlers::feed::feed_delta))
         .route("/api/v1/hashtags", get(handlers::feed::hashtags))
+        .route("/api/v1/uploaders", get(handlers::feed::uploaders))
         // Social
         .route(
             "/api/v1/upload/{id}/like",
@@ -219,45 +245,101 @@ async fn main() -> Result<()> {
         api
     };
 
-    // Serve media files from disk
-    let media_service = ServeDir::new(&config.media_path);
-
+    // NOTE: media is deliberately NOT served over HTTP.
+    //
+    // Files live under `media_path` so the compression worker and the export job can read
+    // them off disk, but nothing may pull them straight from `/media/**` — that bypasses
+    // the visibility checks (soft-delete + ban-hide) that make a host takedown stick.
+    // Every legitimate fetch goes through `/api/v1/upload/{id}/{original,preview,display,
+    // thumbnail}`, which filter via `find_visible_media`; those are the only media URLs the
+    // backend ever emits (see `handlers::feed`).
+    //
+    // This used to be a `ServeDir` on `/media` with four `nest_service` blockers on the
+    // subtrees above it. That was bypassable: axum routes on the RAW path while `ServeDir`
+    // percent-decodes afterwards, so `/media/%70reviews/{id}.jpg` missed every blocker,
+    // fell through to the `ServeDir`, and was decoded back to `previews/` on disk — serving
+    // a taken-down photo to anyone, unauthenticated. Any single escaped byte worked, in all
+    // four subtrees. Deleting the route removes the vector outright rather than racing the
+    // decoder; `/media/**` now 404s regardless of encoding.
     let router = Router::new()
-        .route("/health", get(|| async { "ok" }))
+        // ONE probe, and it touches the database. The merge of the unattended-blockers work
+        // brought a competing design — a dependency-free `/health` for the compose gate plus
+        // a DB-backed `/health/ready` for an external monitor. That split is defensible, and
+        // it was rejected deliberately:
+        //
+        //   * `/health` returning a constant "ok" is the exact defect faea555 fixed and
+        //     verified live (stop Postgres → 503 → start → 200, with no app restart). Every
+        //     request path touches the database, so a constant probe reports healthy while
+        //     the app is useless — the disk-full endgame stayed green all the way down.
+        //   * The split's motive was that Caddy's `depends_on: app: service_healthy` would
+        //     be blocked by a Postgres hiccup at boot. But `app` itself already gates on
+        //     `db: service_healthy`, so the DB is up before this probe ever runs, and the
+        //     healthcheck carries a 20s start_period plus 5 retries on top.
+        //   * The two handlers were the same `SELECT 1` with the same 2s timeout under two
+        //     names, so keeping both bought nothing.
+        //
+        // The external uptime monitor the runbook now calls for points at this route.
+        .route("/health", get(health))
         .merge(api)
-        // Block direct HTTP access to ALL media subtrees. The files live under
-        // `media_path` (so the compression worker and export can read them off disk) but
-        // must NOT be pullable straight from `/media/**` — that bypasses the visibility
-        // checks (soft-delete + ban-hide) in the gated handlers. Every legitimate fetch
-        // goes through `/api/v1/upload/{id}/{original,preview,thumbnail}`, which filter
-        // via `find_visible_media`. The more specific nests take precedence over the
-        // `/media` ServeDir below (which, with all three subtrees blocked, now serves
-        // nothing — kept as a backstop).
-        .nest_service(
-            "/media/originals",
-            get(|| async { axum::http::StatusCode::NOT_FOUND }),
-        )
-        .nest_service(
-            "/media/previews",
-            get(|| async { axum::http::StatusCode::NOT_FOUND }),
-        )
-        .nest_service(
-            "/media/thumbnails",
-            get(|| async { axum::http::StatusCode::NOT_FOUND }),
-        )
-        .nest_service("/media", media_service)
         .layer(TraceLayer::new_for_http())
         .with_state(state);
 
     let listener = tokio::net::TcpListener::bind(("0.0.0.0", config.app_port)).await?;
     tracing::info!("listening on {}", listener.local_addr()?);
-    axum::serve(listener, router)
-        .with_graceful_shutdown(shutdown_signal())
-        .await?;
+    // `into_make_service_with_connect_info` is required by the pre-auth handlers, which
+    // extract `ConnectInfo<SocketAddr>` to use the peer address as the rate-limit key when
+    // X-Forwarded-For is absent. Without it those extractors fail at runtime.
+    axum::serve(
+        listener,
+        router.into_make_service_with_connect_info::<std::net::SocketAddr>(),
+    )
+    .with_graceful_shutdown(shutdown_signal())
+    .await?;
 
     Ok(())
 }
 
+/// How long `/health` waits for the database before calling the app unhealthy. Deliberately
+/// short: the point is to answer "can this process actually serve a request right now", and a
+/// probe that blocks for the acquire timeout is itself a symptom.
+const HEALTH_DB_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(2);
+
+/// Readiness probe — the Docker healthcheck, Caddy's `depends_on` gate, and the runbook's
+/// event-day `curl` all hit this.
+///
+/// It used to return the literal string `"ok"` and touch nothing. Every request in the app needs
+/// the database, so that answered a question nobody asked: the container reported healthy while
+/// every real request 500'd, and with no operator watching during the event there was no signal
+/// at all. Note what this does NOT buy: Compose's `restart: unless-stopped` does not react to
+/// healthcheck state, so nothing restarts on a red probe — this is a diagnostic, and it is
+/// deliberately not wired to automatic recovery, because the pool already heals itself across a
+/// Postgres restart (sqlx revalidates on acquire) and an auto-restart would truncate every
+/// in-flight upload to "fix" an outage that was about to clear on its own.
+async fn health(
+    axum::extract::State(state): axum::extract::State<AppState>,
+) -> impl axum::response::IntoResponse {
+    use axum::http::StatusCode;
+    match tokio::time::timeout(
+        HEALTH_DB_TIMEOUT,
+        sqlx::query("SELECT 1").execute(&state.pool),
+    )
+    .await
+    {
+        Ok(Ok(_)) => (StatusCode::OK, "ok"),
+        Ok(Err(e)) => {
+            tracing::error!(error = ?e, "health check: database query failed");
+            (StatusCode::SERVICE_UNAVAILABLE, "database unavailable")
+        }
+        Err(_) => {
+            tracing::error!(
+                timeout_s = HEALTH_DB_TIMEOUT.as_secs(),
+                "health check: database did not respond"
+            );
+            (StatusCode::SERVICE_UNAVAILABLE, "database timeout")
+        }
+    }
+}
+
 /// Hard cap on how long we wait for in-flight connections to drain after a shutdown
 /// signal. Uploads (streamed to disk) finish in well under this; the cap exists because
 /// long-lived SSE streams never end on their own and would otherwise keep the graceful

backend/src/models/hashtag.rs

@@ -63,6 +63,33 @@ impl Hashtag {
             .await?;
         Ok(())
     }
+
+    /// The upload's current tags, lowercased and sorted — the comparable form.
+    ///
+    /// Exists so `edit_upload` can tell a real hashtag change from a re-send of the same list.
+    /// Without it, `PATCH {"hashtags": []}` in a loop retired the HTML keepsake on every request
+    /// (readiness is derived from `event.export_epoch`), and every armed rebuild was superseded
+    /// before the debounce let it start — so the viewer 404'd for the rest of the event at zero
+    /// cost to the client. One indexed lookup on `upload_hashtag(upload_id)` is a fair price for
+    /// closing that.
+    pub async fn normalized_for_upload<'e, E>(
+        executor: E,
+        upload_id: Uuid,
+    ) -> Result<Vec<String>, sqlx::Error>
+    where
+        E: sqlx::PgExecutor<'e>,
+    {
+        let rows: Vec<(String,)> = sqlx::query_as(
+            "SELECT lower(h.tag) FROM hashtag h
+             JOIN upload_hashtag uh ON uh.hashtag_id = h.id
+             WHERE uh.upload_id = $1
+             ORDER BY lower(h.tag)",
+        )
+        .bind(upload_id)
+        .fetch_all(executor)
+        .await?;
+        Ok(rows.into_iter().map(|(t,)| t).collect())
+    }
 }
 
 /// Extract `#hashtags` from text (caption or body). Tags are restricted to

backend/src/models/upload.rs

@@ -46,12 +46,16 @@ pub struct VisibleMedia {
     pub original_path: String,
     pub preview_path: Option<String>,
     pub thumbnail_path: Option<String>,
+    pub display_path: Option<String>,
     pub mime_type: String,
 }
 
 impl Upload {
     /// Takes any executor so the caller can run it inside a transaction (atomic
     /// quota + insert) or standalone against the pool.
+    // Eight arguments, one per column the INSERT writes, with exactly one call site. A params
+    // struct here would restate the column list a second time and buy nothing.
+    #[allow(clippy::too_many_arguments)]
     pub async fn create<'e, E>(
         executor: E,
         event_id: Uuid,
@@ -60,13 +64,23 @@ impl Upload {
         mime_type: &str,
         original_size_bytes: i64,
         caption: Option<&str>,
-    ) -> Result<Self, sqlx::Error>
+        client_upload_id: Option<Uuid>,
+    ) -> Result<Option<Self>, sqlx::Error>
     where
         E: sqlx::PgExecutor<'e>,
     {
+        // `Ok(None)` means this exact `client_upload_id` is already stored — the caller's request
+        // is a retry of one that already succeeded, and it must replay the original row rather
+        // than create a second. Letting the unique index raise instead would work, but only after
+        // the whole transaction had aborted, and it would arrive as an opaque database error the
+        // caller would have to string-match to recognise.
+        //
+        // The conflict target repeats the index's `WHERE` clause because it is a partial index;
+        // without it Postgres cannot prove which index to use and rejects the statement.
         sqlx::query_as::<_, Self>(
-            "INSERT INTO upload (event_id, user_id, original_path, mime_type, original_size_bytes, caption)
-             VALUES ($1, $2, $3, $4, $5, $6)
+            "INSERT INTO upload (event_id, user_id, original_path, mime_type, original_size_bytes, caption, client_upload_id)
+             VALUES ($1, $2, $3, $4, $5, $6, $7)
+             ON CONFLICT (client_upload_id) WHERE client_upload_id IS NOT NULL DO NOTHING
              RETURNING *",
         )
         .bind(event_id)
@@ -75,7 +89,29 @@ impl Upload {
         .bind(mime_type)
         .bind(original_size_bytes)
         .bind(caption)
-        .fetch_one(executor)
+        .bind(client_upload_id)
+        .fetch_optional(executor)
+        .await
+    }
+
+    /// Look up a live upload by the idempotency key its client sent.
+    ///
+    /// Scoped to the user as well as the key: the key alone is unique, but a lookup that ignored
+    /// ownership would let one guest's retry return another guest's row if a key ever repeated.
+    /// Soft-deleted rows are excluded on purpose — if the guest deleted the photo and their queue
+    /// later retries, they should get a fresh upload rather than a resurrection of a deleted one.
+    pub async fn find_by_client_upload_id(
+        pool: &sqlx::PgPool,
+        user_id: Uuid,
+        client_upload_id: Uuid,
+    ) -> Result<Option<Self>, sqlx::Error> {
+        sqlx::query_as::<_, Self>(
+            "SELECT * FROM upload
+             WHERE client_upload_id = $1 AND user_id = $2 AND deleted_at IS NULL",
+        )
+        .bind(client_upload_id)
+        .bind(user_id)
+        .fetch_optional(pool)
         .await
     }
 
@@ -93,7 +129,7 @@ impl Upload {
         id: Uuid,
     ) -> Result<Option<VisibleMedia>, sqlx::Error> {
         sqlx::query_as::<_, VisibleMedia>(
-            "SELECT up.original_path, up.preview_path, up.thumbnail_path, up.mime_type
+            "SELECT up.original_path, up.preview_path, up.thumbnail_path, up.display_path, up.mime_type
              FROM upload up
              JOIN \"user\" u ON u.id = up.user_id
              WHERE up.id = $1 AND up.deleted_at IS NULL
@@ -134,6 +170,82 @@ impl Upload {
         Ok(())
     }
 
+    pub async fn set_display_path(
+        pool: &PgPool,
+        id: Uuid,
+        display_path: &str,
+    ) -> Result<(), sqlx::Error> {
+        sqlx::query("UPDATE upload SET display_path = $2 WHERE id = $1")
+            .bind(id)
+            .bind(display_path)
+            .execute(pool)
+            .await?;
+        Ok(())
+    }
+
+    /// Stamp which revision of the derivative pipeline produced this row's preview/display,
+    /// so the startup backfill can find rows generated by an older one exactly once.
+    ///
+    /// Also clears the attempt counter: success is the only thing that resets it, and folding
+    /// the reset in here means both the live path and the backfill get it with no extra call
+    /// site to forget.
+    pub async fn set_derivatives_rev(pool: &PgPool, id: Uuid, rev: i16) -> Result<(), sqlx::Error> {
+        sqlx::query(
+            "UPDATE upload
+                SET derivatives_rev = $2, derivative_attempts = 0, derivative_last_error = NULL
+              WHERE id = $1",
+        )
+        .bind(id)
+        .bind(rev)
+        .execute(pool)
+        .await?;
+        Ok(())
+    }
+
+    /// Record that derivative processing is ABOUT to be attempted, returning the new count.
+    ///
+    /// WRITE-AHEAD ON PURPOSE. The failure this bounds is a cgroup SIGKILL: the process
+    /// vanishes mid-work, so no `Err` is returned, no error handler runs and no `Drop` fires.
+    /// A counter incremented after a failure would increment zero times per crash and the
+    /// boot loop would be unchanged. Counting the ATTEMPT is the only thing that survives the
+    /// process dying. The cost is that a genuinely transient failure also burns an attempt —
+    /// acceptable, because the retry budget is per-boot-loop, not per-request, and success
+    /// resets it to zero.
+    /// `None` when the row no longer exists (hard-deleted, or an e2e TRUNCATE landed while the
+    /// task waited on the semaphore) — the caller should abandon quietly rather than treat a
+    /// missing row as a processing failure.
+    pub async fn begin_derivative_attempt(
+        pool: &PgPool,
+        id: Uuid,
+    ) -> Result<Option<i16>, sqlx::Error> {
+        sqlx::query_scalar(
+            "UPDATE upload
+                SET derivative_attempts = derivative_attempts + 1
+              WHERE id = $1
+          RETURNING derivative_attempts",
+        )
+        .bind(id)
+        .fetch_optional(pool)
+        .await
+    }
+
+    /// Store why the last derivative attempt failed. Diagnostics only — nothing branches on it.
+    pub async fn record_derivative_failure(
+        pool: &PgPool,
+        id: Uuid,
+        error: &str,
+    ) -> Result<(), sqlx::Error> {
+        // Bounded: an anyhow chain can be long, and this is written on a failure path that may
+        // repeat across every row of a bad batch.
+        let truncated: String = error.chars().take(500).collect();
+        sqlx::query("UPDATE upload SET derivative_last_error = $2 WHERE id = $1")
+            .bind(id)
+            .bind(truncated)
+            .execute(pool)
+            .await?;
+        Ok(())
+    }
+
     pub async fn set_thumbnail_path(
         pool: &PgPool,
         id: Uuid,
@@ -147,40 +259,8 @@ impl Upload {
         Ok(())
     }
 
-    /// Soft-deletes the upload and decrements the uploader's `total_upload_bytes`.
-    /// Done in a single transaction so a crash between the two writes can't leave
-    /// the quota counter pointing at bytes the user has already deleted (which would
-    /// silently lock them out of future uploads).
-    ///
-    /// No-op if the row is already deleted — protects against a double-tap on the
-    /// delete action double-decrementing the counter.
-    pub async fn soft_delete(pool: &PgPool, id: Uuid) -> Result<(), sqlx::Error> {
-        let mut tx = pool.begin().await?;
-        let row: Option<(Uuid, i64)> = sqlx::query_as(
-            "UPDATE upload
-             SET deleted_at = NOW()
-             WHERE id = $1 AND deleted_at IS NULL
-             RETURNING user_id, original_size_bytes",
-        )
-        .bind(id)
-        .fetch_optional(&mut *tx)
-        .await?;
-        if let Some((user_id, bytes)) = row {
-            sqlx::query(
-                "UPDATE \"user\"
-                 SET total_upload_bytes = GREATEST(0, total_upload_bytes - $2)
-                 WHERE id = $1",
-            )
-            .bind(user_id)
-            .bind(bytes)
-            .execute(&mut *tx)
-            .await?;
-        }
-        tx.commit().await?;
-        Ok(())
-    }
-
-    /// Event-scoped variant of [`Self::soft_delete`]. Returns `false` if no row
+    /// Soft-deletes an upload within its event and refunds the uploader's
+    /// `total_upload_bytes`, in one transaction. Returns `false` if no row
     /// matched (already deleted, wrong event, or unknown id) so host handlers
     /// can return a clean 404 instead of silently no-op'ing.
     /// Executor-generic so a caller can run the delete and the keepsake regeneration in ONE

backend/src/models/user.rs

@@ -60,6 +60,54 @@ impl User {
         .await
     }
 
+    /// Create a user with an explicit role, in ONE statement.
+    ///
+    /// `create` + a separate `UPDATE ... SET role` is not equivalent: a crash or a pool error
+    /// between the two leaves a GUEST row holding a reserved name, which is exactly the
+    /// poisoned state that bricked admin login — now self-inflicted, and invisible to a
+    /// role-based lookup, so the next login would create yet another.
+    pub async fn create_with_role(
+        pool: &PgPool,
+        event_id: Uuid,
+        display_name: &str,
+        pin_hash: &str,
+        role: UserRole,
+    ) -> Result<Self, sqlx::Error> {
+        sqlx::query_as::<_, Self>(
+            "INSERT INTO \"user\" (event_id, display_name, recovery_pin_hash, role)
+             VALUES ($1, $2, $3, $4)
+             RETURNING *",
+        )
+        .bind(event_id)
+        .bind(display_name)
+        .bind(pin_hash)
+        .bind(role)
+        .fetch_one(pool)
+        .await
+    }
+
+    /// The event's admin, looked up BY ROLE.
+    ///
+    /// The name is not the identity and never was. Looking the admin up by `display_name`
+    /// meant any guest who joined as "Admin" first made the lookup miss, and the fallback
+    /// `create` then violated the case-insensitive unique index from migration 007 — a
+    /// permanent 500 on admin login, recoverable only by hand-editing the database.
+    ///
+    /// `ORDER BY created_at` so a database that somehow acquired two admin rows resolves to a
+    /// stable one rather than alternating between them.
+    pub async fn find_admin_for_event(
+        pool: &PgPool,
+        event_id: Uuid,
+    ) -> Result<Option<Self>, sqlx::Error> {
+        sqlx::query_as::<_, Self>(
+            "SELECT * FROM \"user\" WHERE event_id = $1 AND role = 'admin'
+             ORDER BY created_at ASC LIMIT 1",
+        )
+        .bind(event_id)
+        .fetch_optional(pool)
+        .await
+    }
+
     pub async fn find_by_id(pool: &PgPool, id: Uuid) -> Result<Option<Self>, sqlx::Error> {
         sqlx::query_as::<_, Self>("SELECT * FROM \"user\" WHERE id = $1")
             .bind(id)
@@ -96,14 +144,31 @@ impl User {
         Ok(row.0)
     }
 
+    /// Window after which a failed-PIN streak is forgotten. Matches the lockout duration, so
+    /// "wait out the cooldown" and "start clean" are the same interval to a guest.
+    const PIN_ATTEMPT_DECAY_MINUTES: i64 = 15;
+
+    /// Record a wrong PIN and return the CURRENT streak length.
+    ///
+    /// The counter decays: before this, it only ever cleared on a successful recovery or after
+    /// a lockout expired, so ordinary typos accumulated across days and a guest could arrive at
+    /// an event already most of the way to being locked out by mistakes made the night before.
+    /// Decay is what makes the raised lock threshold safe rather than merely lenient.
     pub async fn increment_failed_pin(pool: &PgPool, id: Uuid) -> Result<i16, sqlx::Error> {
         let row: (i16,) = sqlx::query_as(
             "UPDATE \"user\"
-             SET failed_pin_attempts = failed_pin_attempts + 1
+             SET failed_pin_attempts = CASE
+                     WHEN last_failed_pin_at IS NULL
+                       OR last_failed_pin_at < NOW() - ($2 || ' minutes')::interval
+                     THEN 1
+                     ELSE failed_pin_attempts + 1
+                 END,
+                 last_failed_pin_at = NOW()
              WHERE id = $1
              RETURNING failed_pin_attempts",
         )
         .bind(id)
+        .bind(Self::PIN_ATTEMPT_DECAY_MINUTES.to_string())
         .fetch_one(pool)
         .await?;
         Ok(row.0)
@@ -124,7 +189,9 @@ impl User {
 
     pub async fn reset_pin_attempts(pool: &PgPool, id: Uuid) -> Result<(), sqlx::Error> {
         sqlx::query(
-            "UPDATE \"user\" SET failed_pin_attempts = 0, pin_locked_until = NULL WHERE id = $1",
+            "UPDATE \"user\"
+                SET failed_pin_attempts = 0, pin_locked_until = NULL, last_failed_pin_at = NULL
+              WHERE id = $1",
         )
         .bind(id)
         .execute(pool)

backend/src/services/compression.rs

@@ -48,6 +48,31 @@ impl CompressionWorker {
         self.generation.fetch_add(1, Ordering::SeqCst);
     }
 
+    /// How many times `do_process` is attempted before an upload is given up on. The
+    /// give-up path is user-visible (the photo disappears), so transient infrastructure
+    /// errors must not reach it.
+    const MAX_PROCESS_ATTEMPTS: u32 = 3;
+
+    /// Revision of the image-derivative pipeline. Bump this whenever a change makes existing
+    /// previews/displays wrong, so `backfill_stale_derivatives` regenerates them once on the
+    /// next start. Rev 1 = EXIF orientation is applied.
+    const DERIVATIVES_REV: i16 = 1;
+
+    /// How many times derivative generation may be ATTEMPTED for one upload before it is left
+    /// alone. Counted write-ahead and reset on success — see `Upload::begin_derivative_attempt`.
+    ///
+    /// This is what turns a fatal input from an outage into a blemish. The startup backfill
+    /// runs unconditionally on every boot, so before this bound a row whose processing killed
+    /// the process was re-selected and re-run forever, and `restart: unless-stopped` made that
+    /// an infinite loop that also dropped every SSE stream and truncated every in-flight
+    /// upload on each cycle. Three attempts absorbs genuinely transient infrastructure
+    /// failures (an ENOSPC spike, a pool blip) without ever becoming unbounded.
+    const MAX_DERIVATIVE_ATTEMPTS: i16 = 3;
+
+    /// Rows regenerated per boot. Bounds both the query and the amount of work a single start
+    /// can queue; whatever is left is picked up on the next boot.
+    const BACKFILL_BATCH: i64 = 200;
+
     /// Spawn a background task to process an uploaded file.
     pub fn process(&self, upload_id: Uuid, original_path: String, mime_type: String) {
         let worker = self.clone();
@@ -60,10 +85,43 @@ impl CompressionWorker {
             if worker.generation.load(Ordering::SeqCst) != born_at {
                 return;
             }
-            match worker
-                .do_process(upload_id, &original_path, &mime_type)
-                .await
-            {
+            // Retry before giving up. Most failures here are transient and self-clearing —
+            // an ENOSPC spike while several guests upload at once, a momentary DB-pool
+            // exhaustion, a panic inside the image codec — and the give-up path is
+            // user-visible data loss, so it is worth a few seconds to avoid entering it.
+            //
+            // But only for failures that CAN clear. An image that exceeds the decode budget,
+            // is corrupt, or is in an unsupported format fails identically on every attempt,
+            // so retrying it just burns 2s + 4s of backoff and writes three near-identical
+            // warnings before reaching the same conclusion. Give up on those immediately.
+            let mut attempt = 1u32;
+            let outcome = loop {
+                match worker
+                    .do_process(upload_id, &original_path, &mime_type)
+                    .await
+                {
+                    Ok(v) => break Ok(v),
+                    Err(e)
+                        if attempt < Self::MAX_PROCESS_ATTEMPTS
+                            && !crate::services::imaging::is_permanent_image_error(&e)
+                            && !crate::services::imaging::is_storage_full_error(&e) =>
+                    {
+                        tracing::warn!(
+                            error = ?e, %upload_id, attempt,
+                            "compression attempt failed; retrying"
+                        );
+                        tokio::time::sleep(std::time::Duration::from_secs(2u64.pow(attempt))).await;
+                        attempt += 1;
+                        // The data may have been reset while we slept (e2e TRUNCATE).
+                        if worker.generation.load(Ordering::SeqCst) != born_at {
+                            return;
+                        }
+                    }
+                    Err(e) => break Err(e),
+                }
+            };
+
+            match outcome {
                 Ok(_) => {
                     tracing::info!("compression completed for upload {upload_id}");
                     let _ = worker.sse_tx.send(SseEvent {
@@ -71,29 +129,75 @@ impl CompressionWorker {
                         data: serde_json::json!({ "upload_id": upload_id }).to_string(),
                     });
                 }
-                Err(e) => {
-                    tracing::error!("compression failed for upload {upload_id}: {e:#}");
-                    // Auto-cleanup: a failed transcode would otherwise leave a
-                    // permanently broken feed card, silently charge the uploader's
-                    // quota, and orphan the original on disk. Refund + soft-delete
-                    // (one tx, so v_feed excludes it), remove the orphan file, then
-                    // tell the uploader (upload-error toast) and evict the card
-                    // everywhere (upload-deleted, already handled by the feed).
+                Err(e) if crate::services::imaging::is_storage_full_error(&e) => {
+                    // Out of disk. Keep the row AND the original — the opposite of the branch
+                    // below, and for the same reason it retains the file: nothing here is the
+                    // guest's fault and nothing about the photo is wrong.
+                    //
+                    // Soft-deleting on ENOSPC was strictly harmful. It refunded the quota while
+                    // keeping the bytes, so it freed nothing, removed the photo from the feed
+                    // seconds after a `201 Created`, and handed the guest the allowance to
+                    // upload it again into the same full disk. Leaving the row live costs
+                    // nothing instead: every client already falls back to the original when
+                    // `preview_url` and `thumbnail_url` are NULL, so the photo stays visible —
+                    // just uncompressed — and `backfill_stale_derivatives` regenerates the
+                    // derivatives on the next start, once there is room for them.
+                    tracing::error!(
+                        %upload_id,
+                        "compression failed: the media filesystem is out of space. The upload is \
+                         kept and served from its original; free disk space and restart to \
+                         regenerate derivatives: {e:#}"
+                    );
                     let _ = Upload::set_compression_status(&worker.pool, upload_id, "failed").await;
-                    if let Err(del) = Upload::soft_delete(&worker.pool, upload_id).await {
-                        tracing::warn!(error = ?del, %upload_id, "failed to soft-delete after compression failure");
-                    }
-                    let orphan = worker.media_path.join(&original_path);
-                    if let Err(rm) = tokio::fs::remove_file(&orphan).await {
-                        tracing::warn!(error = ?rm, path = %orphan.display(), "failed to remove orphaned original");
-                    }
+                    // Not an "error" event: nothing was lost and there is nothing for the guest
+                    // to act on. Clients treat this purely as "refetch me", which is what makes
+                    // the card appear with its original as the image source.
+                    let _ = worker.sse_tx.send(SseEvent {
+                        event_type: "upload-processed".to_string(),
+                        data: serde_json::json!({ "upload_id": upload_id }).to_string(),
+                    });
+                }
+                Err(e) => {
+                    tracing::error!(
+                        "compression failed for upload {upload_id} after {attempt} attempt(s): {e:#}"
+                    );
+                    // KEEP THE ROW. This used to soft-delete, which made a derivative failure
+                    // indistinguishable — to the guest — from their photo being deleted: they
+                    // got a `201 Created`, watched the card appear, and then watched it vanish.
+                    // The row left `v_feed`, `find_visible_media` and BOTH keepsake archives,
+                    // so the photo was gone from the product's core promise while its bytes sat
+                    // on disk for 14 days waiting for a `cleanup_deleted_media` that nothing
+                    // told anyone about. There is no host or admin screen listing compression
+                    // failures, so recovery meant hand-written SQL that also had to re-add the
+                    // refunded quota bytes. Against "0 lost uploads", that was silent per-photo
+                    // loss on any error the ENOSPC arm above doesn't catch — a HEIC that slipped
+                    // the allowlist, a truncated frame, an ffmpeg hiccup, a pool blip.
+                    //
+                    // This is exactly what the ENOSPC arm already does and documents as correct:
+                    // every client falls back to the original when `preview_url` and
+                    // `thumbnail_url` are NULL, so the photo stays visible and downloadable —
+                    // just uncompressed — and `backfill_stale_derivatives` retries it on the
+                    // next boot, now bounded by `derivative_attempts` so a poisoned row cannot
+                    // loop. The quota stays charged, which is correct: the bytes are still on
+                    // disk and still the guest's.
+                    let _ = Upload::set_compression_status(&worker.pool, upload_id, "failed").await;
+                    tracing::warn!(
+                        %upload_id,
+                        path = %worker.media_path.join(&original_path).display(),
+                        "derivatives failed; the upload is kept and served from its original"
+                    );
+                    // `upload-error` still fires so the uploader learns the photo will look
+                    // uncompressed. `upload-deleted` deliberately does NOT — nothing was
+                    // deleted, and evicting the card was the visible half of the data loss.
                     let _ = worker.sse_tx.send(SseEvent {
                         event_type: "upload-error".to_string(),
                         data: serde_json::json!({ "upload_id": upload_id, "error": e.to_string() })
                             .to_string(),
                     });
+                    // Tell every client to refetch, so the card re-renders from the original
+                    // instead of sitting on a stale "processing" placeholder forever.
                     let _ = worker.sse_tx.send(SseEvent {
-                        event_type: "upload-deleted".to_string(),
+                        event_type: "upload-processed".to_string(),
                         data: serde_json::json!({ "upload_id": upload_id }).to_string(),
                     });
                 }
@@ -112,129 +216,632 @@ impl CompressionWorker {
         let original = self.media_path.join(original_path);
 
         if mime_type.starts_with("image/") {
-            let preview_rel = self
-                .generate_image_preview(upload_id, &original, mime_type)
+            // Count the attempt BEFORE doing the work — see `begin_derivative_attempt`. If this
+            // input is the one that kills the container, this write is the only record that
+            // survives, and it is what stops the boot backfill replaying it forever.
+            match Upload::begin_derivative_attempt(&self.pool, upload_id).await? {
+                Some(attempts) if attempts > Self::MAX_DERIVATIVE_ATTEMPTS => {
+                    anyhow::bail!(
+                        "derivative generation gave up after {} attempt(s)",
+                        attempts - 1
+                    );
+                }
+                Some(_) => {}
+                // The row vanished while this task waited on the semaphore. Nothing to do, and
+                // reporting a failure would broadcast into a stream that no longer has a card.
+                None => return Ok(()),
+            }
+            let (preview_rel, display_rel) = self
+                .generate_image_derivatives(upload_id, &original, mime_type)
                 .await?;
             Upload::set_preview_path(&self.pool, upload_id, &preview_rel).await?;
-            tracing::info!("preview generated for upload {upload_id}");
+            Upload::set_display_path(&self.pool, upload_id, &display_rel).await?;
+            Upload::set_derivatives_rev(&self.pool, upload_id, Self::DERIVATIVES_REV).await?;
+            tracing::info!("preview + display generated for upload {upload_id}");
         } else if mime_type.starts_with("video/") {
-            let thumb_rel = self.generate_video_thumbnail(upload_id, &original).await?;
-            Upload::set_thumbnail_path(&self.pool, upload_id, &thumb_rel).await?;
-            tracing::info!("thumbnail generated for upload {upload_id}");
+            // A missing poster must NOT fail the upload. `set_thumbnail_path` is only reached when
+            // a file really exists, so `thumbnail_path` stays NULL otherwise — which every consumer
+            // already handles (FeedListCard, VirtualFeed, LightboxModal are all null-safe).
+            //
+            // The `?` here used to hide the defect; making the check strict without also making
+            // this non-fatal would have been far worse than the bug. Every clip of a second or less
+            // would fail compression, exhaust its retries and be soft-deleted — a cosmetic defect
+            // turned into data loss, on exactly the mis-tap/Live-Photo clips guests produce most.
+            // Handling only the `Ok(None)` arm was not enough: the `?` on the call itself still
+            // routed every OTHER poster failure into the give-up path. `extract_poster_frame`
+            // returns `Err` when ffmpeg is missing from the image, when it hangs on a truncated
+            // `.mov` and trips FFMPEG_TIMEOUT, or when `thumbnails/` can't be created — and
+            // `set_thumbnail_path` returns `Err` on any DB blip. None of those say anything about
+            // the video itself, yet each one destroyed it. Confirmed live: on a box with no ffmpeg
+            // the spawn error propagated, exhausted all three attempts and soft-deleted the clip.
+            //
+            // Nothing about a video post depends on the poster — `get_original` serves the file
+            // byte-for-byte and the tile falls back to the video element — so no failure in this
+            // branch may fail the upload.
+            match self.generate_video_thumbnail(upload_id, &original).await {
+                Ok(Some(thumb_rel)) => {
+                    match Upload::set_thumbnail_path(&self.pool, upload_id, &thumb_rel).await {
+                        Ok(()) => tracing::info!("thumbnail generated for upload {upload_id}"),
+                        Err(e) => tracing::warn!(
+                            error = ?e, %upload_id,
+                            "poster extracted but could not be recorded; the video keeps its own tile"
+                        ),
+                    }
+                }
+                Ok(None) => {
+                    tracing::warn!(
+                        %upload_id,
+                        "no poster frame could be extracted; the video keeps its own tile"
+                    );
+                }
+                Err(e) => {
+                    tracing::warn!(
+                        error = ?e, %upload_id,
+                        "poster extraction failed; the video keeps its own tile"
+                    );
+                }
+            }
         }
 
         Upload::set_compression_status(&self.pool, upload_id, "done").await?;
         Ok(())
     }
 
-    async fn generate_image_preview(
+    /// Longest edge of the big-screen "display" derivative used by the diashow. Sized to be
+    /// sharp on 1080p/4K while staying bounded (a ~2048px JPEG decodes to ~16 MB — trivial
+    /// for any kiosk, unlike a raw multi-thousand-pixel original).
+    const DISPLAY_MAX_EDGE: u32 = 2048;
+    /// Longest edge of the phone-feed "preview" (data-saver default).
+    const PREVIEW_MAX_EDGE: u32 = 800;
+
+    /// Above this pixel count the PNG original is stored as uploaded, unoptimised.
+    ///
+    /// oxipng's peak memory scales with PIXELS, not file size: it decodes the PNG itself and
+    /// then evaluates row filters, each trial holding a full-size buffer. That is why a 2.82
+    /// MiB file could measure 1250 MiB of peak RSS inside a 1 GiB container — smooth,
+    /// synthetic content compresses to almost nothing on disk while still being 8000x8000.
+    /// 8 MP covers every real phone photo; beyond it we decline the (lossless, cosmetic)
+    /// saving rather than risk the OOM kill.
+    const OXIPNG_MAX_PIXELS: u64 = 8_000_000;
+
+
+    /// Wall-clock ceiling for one oxipng run.
+    ///
+    /// Bounds TIME, NOT MEMORY — oxipng checks the deadline between trials, so a single trial
+    /// still allocates in full. The pixel gate above and the sequential build (see
+    /// `default-features = false` in Cargo.toml) are what bound memory. Do not treat this
+    /// constant as the OOM fix.
+    const OXIPNG_TIMEOUT: std::time::Duration = std::time::Duration::from_secs(20);
+
+    /// Decode the image ONCE and emit both derivatives — the 800px `preview` (phone feed)
+    /// and the 2048px `display` (diashow). Returns `(preview_rel, display_rel)`.
+    async fn generate_image_derivatives(
         &self,
         upload_id: Uuid,
         original: &Path,
         mime_type: &str,
-    ) -> Result<String> {
+    ) -> Result<(String, String)> {
         let previews_dir = self.media_path.join("previews");
+        let displays_dir = self.media_path.join("displays");
         tokio::fs::create_dir_all(&previews_dir).await?;
+        tokio::fs::create_dir_all(&displays_dir).await?;
 
-        let preview_filename = format!("{upload_id}.jpg");
-        let preview_path = previews_dir.join(&preview_filename);
+        let filename = format!("{upload_id}.jpg");
+        let preview_path = previews_dir.join(&filename);
+        let display_path = displays_dir.join(&filename);
         let original = original.to_path_buf();
-        let preview_path_clone = preview_path.clone();
         let mime_owned = mime_type.to_string();
 
-        // Run blocking image operations in a spawn_blocking task
-        tokio::task::spawn_blocking(move || -> Result<()> {
-            // Reject decompression bombs *before* fully decoding: the upload body
-            // cap bounds the file size on disk, but a small file can still decode to
-            // enormous dimensions (e.g. a ~1 MB image expanding to 50k×50k px →
-            // gigabytes), OOM-ing the box during decode/resize. 12000×12000 covers
-            // any real phone photo; max_alloc hard-caps the decode allocation.
-            let mut reader = image::ImageReader::open(&original)
-                .context("failed to open image")?
-                .with_guessed_format()
-                .context("failed to read image header")?;
-            let mut limits = image::Limits::default();
-            limits.max_image_width = Some(12_000);
-            limits.max_image_height = Some(12_000);
-            limits.max_alloc = Some(256 * 1024 * 1024);
-            reader.limits(limits);
-            let img = reader.decode().context("failed to decode image")?;
-
-            // Resize to max 800px wide, preserving aspect ratio
-            let preview = img.resize(800, 800, image::imageops::FilterType::Lanczos3);
-            preview
-                .save_with_format(&preview_path_clone, image::ImageFormat::Jpeg)
-                .context("failed to save preview")?;
-
-            // If the original is PNG, try lossless compression in-place
-            if mime_owned == "image/png" {
-                let opts = oxipng::Options::from_preset(2);
-                let _ = oxipng::optimize(
-                    &oxipng::InFile::Path(original),
-                    &oxipng::OutFile::Path {
-                        path: None,
-                        preserve_attrs: true,
-                    },
-                    &opts,
+        // Estimate the peak from the HEADER (no pixels decoded — the same kind of cheap probe
+        // the upload handler already does via `exceeds_decode_budget`) and, if this job is a
+        // giant, take the exclusive permit so it cannot overlap another giant. Held for the
+        // whole blocking section, released on drop including on error.
+        let estimate =
+            crate::services::imaging::estimated_processing_peak_bytes(&original, Self::DISPLAY_MAX_EDGE);
+        let _heavy_permit = match estimate {
+            Some(bytes) if bytes > crate::services::imaging::HEAVY_IMAGE_BYTES => {
+                tracing::debug!(
+                    %upload_id,
+                    estimated_mib = bytes / (1024 * 1024),
+                    "waiting for the heavy-image permit"
                 );
+                Some(crate::services::imaging::HEAVY_IMAGE_PERMITS.acquire().await)
             }
+            _ => None,
+        };
 
-            Ok(())
+        // Run blocking image operations in a spawn_blocking task
+        tokio::task::spawn_blocking(move || {
+            write_image_derivatives(upload_id, &original, &mime_owned, &preview_path, &display_path)
         })
         .await??;
 
-        Ok(format!("previews/{preview_filename}"))
+        Ok((
+            format!("previews/{filename}"),
+            format!("displays/{filename}"),
+        ))
     }
 
-    async fn generate_video_thumbnail(&self, upload_id: Uuid, original: &Path) -> Result<String> {
+    /// Regenerate image derivatives that an older pipeline produced. Fire-and-forget from
+    /// startup; picks up two cases, both of which leave the ORIGINAL untouched:
+    ///
+    /// - uploads processed before the `display` derivative existed (preview but no
+    ///   `display_path`), and
+    /// - uploads whose derivatives predate `DERIVATIVES_REV` — currently rev 1, which applies
+    ///   the EXIF orientation. Everything generated before it is stored sideways for any
+    ///   portrait phone photo.
+    ///
+    /// Unlike the failure path in `process`, a backfill error is logged and skipped — it must
+    /// NEVER destroy or soft-delete an upload that already has a working preview.
+    ///
+    /// Bounded in three ways, all of them load-bearing on a box that restarts itself:
+    /// `derivative_attempts` stops a fatal row being replayed on every boot, `BACKFILL_BATCH`
+    /// stops one start queueing unbounded work, and the whole thing runs as ONE task walking
+    /// the rows sequentially rather than N tasks racing for the same semaphore.
+    pub async fn backfill_stale_derivatives(&self) {
+        // `original_path IS NOT NULL` was dead — the column is NOT NULL. What actually needs
+        // excluding is the blanked path `cleanup_deleted_media` leaves behind.
+        let rows = sqlx::query_as::<_, (Uuid, String, String)>(
+            "SELECT id, original_path, mime_type FROM upload
+             WHERE deleted_at IS NULL AND mime_type LIKE 'image/%'
+               AND original_path <> ''
+               AND derivative_attempts < $2
+               AND (
+                     (display_path IS NULL AND preview_path IS NOT NULL)
+                  OR derivatives_rev < $1
+               )
+             ORDER BY created_at DESC
+             LIMIT $3",
+        )
+        .bind(Self::DERIVATIVES_REV)
+        .bind(Self::MAX_DERIVATIVE_ATTEMPTS)
+        .bind(Self::BACKFILL_BATCH)
+        .fetch_all(&self.pool)
+        .await;
+        let rows = match rows {
+            Ok(r) => r,
+            Err(e) => {
+                tracing::warn!(error = ?e, "derivative backfill query failed");
+                return;
+            }
+        };
+
+        self.report_exhausted_derivatives().await;
+
+        if rows.is_empty() {
+            return;
+        }
+        tracing::info!("regenerating derivatives for {} upload(s)", rows.len());
+
+        // ONE task for the whole batch. The previous shape spawned a task per row, so a large
+        // backlog created thousands of live tasks that each held a pool handle and queued on
+        // the same two semaphore permits, competing with live uploads for the entire boot.
+        let worker = self.clone();
+        tokio::spawn(async move {
+            for (id, original_path, mime_type) in rows {
+                let _permit = worker.semaphore.acquire().await;
+                // Write-ahead, exactly as in the live path: if this row is the one that kills
+                // the process, this increment is the only thing that outlives the SIGKILL.
+                match Upload::begin_derivative_attempt(&worker.pool, id).await {
+                    Ok(Some(n)) if n > Self::MAX_DERIVATIVE_ATTEMPTS => continue,
+                    Ok(Some(_)) => {}
+                    Ok(None) => continue,
+                    Err(e) => {
+                        tracing::warn!(error = ?e, %id, "could not record a backfill attempt; skipping");
+                        continue;
+                    }
+                }
+                let original = worker.media_path.join(&original_path);
+                match worker
+                    .generate_image_derivatives(id, &original, &mime_type)
+                    .await
+                {
+                    Ok((preview_rel, display_rel)) => {
+                        let _ = Upload::set_preview_path(&worker.pool, id, &preview_rel).await;
+                        let _ = Upload::set_display_path(&worker.pool, id, &display_rel).await;
+                        // Clears derivative_attempts too, so a row that failed transiently is
+                        // not one boot closer to being abandoned.
+                        let _ =
+                            Upload::set_derivatives_rev(&worker.pool, id, Self::DERIVATIVES_REV)
+                                .await;
+                        tracing::info!("derivatives regenerated for upload {id}");
+                    }
+                    Err(e) => {
+                        // Leave the existing derivatives and the original intact; this row is
+                        // retried on the next start until its attempt budget runs out. The rev
+                        // stays behind, which is the marker that it still needs doing.
+                        tracing::warn!(error = ?e, %id, "derivative backfill failed; leaving as-is");
+                        let _ =
+                            Upload::record_derivative_failure(&worker.pool, id, &format!("{e:#}"))
+                                .await;
+                    }
+                }
+            }
+        });
+    }
+
+    /// Re-extract poster frames for videos that never got one.
+    ///
+    /// A video interrupted by a restart is stranded: `startup_recovery` flips its
+    /// `compression_status` from `processing` to `failed` and nothing re-enqueues it, so
+    /// `thumbnail_path` stays NULL forever while the clip itself plays fine. The feed shows a
+    /// posterless tile for the rest of the event, and after
+    /// `FAILED_ORIGINAL_RETENTION_DAYS` the reclaim sweep is entitled to the original.
+    ///
+    /// Shares `derivative_attempts` with the image backfill on purpose. Note the consequence,
+    /// which is intended rather than a bug to fix later: `extract_poster_frame` returning
+    /// `Ok(false)` is a NORMAL, permanent outcome for a sub-second clip (Live Photos,
+    /// mis-taps), and since the counter is write-ahead and only cleared by a real success,
+    /// those clips stop being re-ffmpeg'd on every boot once the budget is spent.
+    pub async fn backfill_video_posters(&self) {
+        let rows = sqlx::query_as::<_, (Uuid, String)>(
+            "SELECT id, original_path FROM upload
+             WHERE deleted_at IS NULL AND mime_type LIKE 'video/%'
+               AND thumbnail_path IS NULL
+               AND original_path <> ''
+               AND derivative_attempts < $1
+             ORDER BY created_at DESC
+             LIMIT $2",
+        )
+        .bind(Self::MAX_DERIVATIVE_ATTEMPTS)
+        .bind(Self::BACKFILL_BATCH)
+        .fetch_all(&self.pool)
+        .await;
+        let rows = match rows {
+            Ok(r) => r,
+            Err(e) => {
+                tracing::warn!(error = ?e, "video poster backfill query failed");
+                return;
+            }
+        };
+        if rows.is_empty() {
+            return;
+        }
+        tracing::info!("re-extracting posters for {} video(s)", rows.len());
+
+        let worker = self.clone();
+        tokio::spawn(async move {
+            for (id, original_path) in rows {
+                let _permit = worker.semaphore.acquire().await;
+                match Upload::begin_derivative_attempt(&worker.pool, id).await {
+                    Ok(Some(n)) if n > Self::MAX_DERIVATIVE_ATTEMPTS => continue,
+                    Ok(Some(_)) => {}
+                    Ok(None) => continue,
+                    Err(e) => {
+                        tracing::warn!(error = ?e, %id, "could not record a poster attempt; skipping");
+                        continue;
+                    }
+                }
+                let original = worker.media_path.join(&original_path);
+                match worker.generate_video_thumbnail(id, &original).await {
+                    Ok(Some(thumb_rel)) => {
+                        if Upload::set_thumbnail_path(&worker.pool, id, &thumb_rel)
+                            .await
+                            .is_ok()
+                        {
+                            // Clears the attempt counter: a video that eventually succeeded
+                            // must not carry a budget scar into a future pipeline revision.
+                            let _ = Upload::set_derivatives_rev(
+                                &worker.pool,
+                                id,
+                                Self::DERIVATIVES_REV,
+                            )
+                            .await;
+                            tracing::info!("poster regenerated for upload {id}");
+                        }
+                    }
+                    // No frame at all — normal for a very short clip. The tile stays
+                    // posterless and the attempt is spent, which is what stops the retry.
+                    Ok(None) => {
+                        tracing::debug!(%id, "still no poster frame; leaving the tile as-is");
+                    }
+                    Err(e) => {
+                        tracing::warn!(error = ?e, %id, "poster backfill failed; leaving as-is");
+                        let _ =
+                            Upload::record_derivative_failure(&worker.pool, id, &format!("{e:#}"))
+                                .await;
+                    }
+                }
+            }
+        });
+    }
+
+    /// Say out loud, once per boot, that some uploads have stopped being retried.
+    ///
+    /// Without this the give-up is invisible: the loop stops (which is the point) but the
+    /// affected photos keep a stale or missing derivative forever with nothing to notice. The
+    /// originals are untouched, so this is recoverable once the cause is fixed — reset
+    /// `derivative_attempts` to 0 and restart.
+    async fn report_exhausted_derivatives(&self) {
+        let exhausted: Result<i64, _> = sqlx::query_scalar(
+            "SELECT count(*) FROM upload
+             WHERE deleted_at IS NULL AND mime_type LIKE 'image/%'
+               AND derivative_attempts >= $2
+               AND (
+                     (display_path IS NULL AND preview_path IS NOT NULL)
+                  OR derivatives_rev < $1
+               )",
+        )
+        .bind(Self::DERIVATIVES_REV)
+        .bind(Self::MAX_DERIVATIVE_ATTEMPTS)
+        .fetch_one(&self.pool)
+        .await;
+        if let Ok(count) = exhausted
+            && count > 0
+        {
+            tracing::error!(
+                count,
+                "{count} upload(s) exhausted derivative regeneration and will no longer be \
+                 retried; their originals are intact — see upload.derivative_last_error, fix \
+                 the cause, then reset derivative_attempts to 0 and restart"
+            );
+        }
+    }
+
+    /// Extract the feed poster for a video. `Ok(None)` when the clip yields no frame — see
+    /// [`crate::services::video::extract_poster_frame`], which owns the seek order, the timeout and
+    /// the artifact check that this function used to be missing.
+    async fn generate_video_thumbnail(
+        &self,
+        upload_id: Uuid,
+        original: &Path,
+    ) -> Result<Option<String>> {
         let thumbs_dir = self.media_path.join("thumbnails");
         tokio::fs::create_dir_all(&thumbs_dir).await?;
 
         let thumb_filename = format!("{upload_id}.jpg");
         let thumb_path = thumbs_dir.join(&thumb_filename);
 
-        // Hard timeout — a malformed video can hang `ffmpeg` indefinitely. Without a
-        // cap, the held compression-worker semaphore permit is never released and the
-        // pool eventually deadlocks (no further uploads ever processed). 120s is well
-        // above the time to extract one frame from any sane input.
-        let mut child = tokio::process::Command::new("ffmpeg")
-            .args([
-                "-i",
-                original.to_str().unwrap_or_default(),
-                "-vframes",
-                "1",
-                "-ss",
-                "00:00:01",
-                "-vf",
-                "scale=800:-1",
-                "-y",
-                thumb_path.to_str().unwrap_or_default(),
-            ])
-            .stdout(std::process::Stdio::piped())
-            .stderr(std::process::Stdio::piped())
-            .kill_on_drop(true)
-            .spawn()
-            .context("failed to spawn ffmpeg")?;
+        let produced =
+            crate::services::video::extract_poster_frame(original, &thumb_path, 800).await?;
 
-        let status =
-            match tokio::time::timeout(std::time::Duration::from_secs(120), child.wait()).await {
-                Ok(res) => res.context("ffmpeg wait failed")?,
-                Err(_) => {
-                    let _ = child.kill().await;
-                    anyhow::bail!("ffmpeg timeout after 120s");
-                }
-            };
-
-        if !status.success() {
-            // Best-effort: drain stderr for the log.
-            let mut stderr = Vec::new();
-            if let Some(mut handle) = child.stderr.take() {
-                use tokio::io::AsyncReadExt;
-                let _ = handle.read_to_end(&mut stderr).await;
-            }
-            anyhow::bail!("ffmpeg failed: {}", String::from_utf8_lossy(&stderr));
-        }
-
-        Ok(format!("thumbnails/{thumb_filename}"))
+        Ok(produced.then(|| format!("thumbnails/{thumb_filename}")))
+    }
+}
+
+/// The blocking half of [`CompressionWorker::generate_image_derivatives`]: decode once, write
+/// both derivatives, then optionally shrink a PNG original in place.
+///
+/// A free function rather than an inline closure so its memory behaviour is directly testable —
+/// this is the code path that OOM-killed the container, and the fix is a scoping property that a
+/// future edit could silently undo.
+fn write_image_derivatives(
+    upload_id: Uuid,
+    original: &Path,
+    mime_type: &str,
+    preview_path: &Path,
+    display_path: &Path,
+) -> Result<()> {
+    let preview_max = CompressionWorker::PREVIEW_MAX_EDGE;
+    let display_max = CompressionWorker::DISPLAY_MAX_EDGE;
+
+    // THE FULL-SIZE DECODE IS SCOPED TO THIS BLOCK ON PURPOSE, and the block yields the
+    // DISPLAY derivative rather than the original.
+    //
+    // `img` is up to 256 MiB (imaging::decode_limits max_alloc) and `resize` only BORROWS it,
+    // so it used to stay alive through both resizes AND the oxipng call below — which decodes
+    // the PNG a second time and holds a full-size buffer per filter trial. That measured
+    // ~1250 MiB of peak RSS for a 2.8 MiB input, inside a 1 GiB cgroup: the container was
+    // SIGKILLed, taking every SSE stream and every in-flight upload with it.
+    //
+    // A block rather than a bare `drop(img)` because a `drop` call is one careless edit away
+    // from being removed as redundant-looking — and note the `else` arm MOVES `img` out, which
+    // is what makes "the block's value is the only survivor" true in both arms.
+    let (display, width, height) = {
+        // Decompression-bomb limits + EXIF orientation, both in one place — see
+        // services::imaging for why neither may be skipped.
+        let img = crate::services::imaging::decode_oriented(original)?;
+        let (width, height) = (img.width(), img.height());
+
+        // Display: max 2048px for the diashow. Only DOWNSCALE — never upscale a smaller
+        // original (that adds bytes with no quality gain); re-encode it as JPEG as-is.
+        let display = if width > display_max || height > display_max {
+            img.resize(
+                display_max,
+                display_max,
+                image::imageops::FilterType::Lanczos3,
+            )
+        } else {
+            img
+        };
+        (display, width, height)
+    };
+
+    display
+        .save_with_format(display_path, image::ImageFormat::Jpeg)
+        .context("failed to save display")?;
+
+    // Preview: max 800px, derived from the DISPLAY, not from the original.
+    //
+    // Both derivatives used to resize the full-size decode independently, so a 8000x8000
+    // original paid for two full-size Lanczos passes and their intermediates — measured 520
+    // MiB peak even after the scoping fix above, which two concurrent workers cannot fit in a
+    // 1 GiB container. Chaining 8000 -> 2048 -> 800 makes the second pass operate on 2048px
+    // input, and the full-size buffer is already freed by the time it runs. Quality is not the
+    // trade-off here: a staged Lanczos3 downscale to 800px is visually indistinguishable from
+    // a single-step one (and is a standard technique for large ratios).
+    display
+        .resize(
+            preview_max,
+            preview_max,
+            image::imageops::FilterType::Lanczos3,
+        )
+        .save_with_format(preview_path, image::ImageFormat::Jpeg)
+        .context("failed to save preview")?;
+    drop(display);
+
+    let pixels = u64::from(width) * u64::from(height);
+
+    // If the original is PNG, try lossless compression in place — but only when its pixel count
+    // is inside the budget, and never for longer than OXIPNG_TIMEOUT. This is a best-effort size
+    // saving: declining it costs disk, while attempting it unbounded cost the whole container.
+    if mime_type == "image/png" {
+        if pixels <= CompressionWorker::OXIPNG_MAX_PIXELS {
+            let mut opts = oxipng::Options::from_preset(2);
+            opts.timeout = Some(CompressionWorker::OXIPNG_TIMEOUT);
+            let _ = oxipng::optimize(
+                &oxipng::InFile::Path(original.to_path_buf()),
+                &oxipng::OutFile::Path {
+                    path: None,
+                    preserve_attrs: true,
+                },
+                &opts,
+            );
+        } else {
+            tracing::info!(
+                %upload_id, pixels,
+                "skipping oxipng: above the pixel budget; the original is stored as uploaded"
+            );
+        }
+    }
+
+    Ok(())
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Peak resident set of THIS process, in bytes, from `/proc/self/status`.
+    fn peak_rss_bytes() -> u64 {
+        let status = std::fs::read_to_string("/proc/self/status").expect("procfs");
+        let line = status
+            .lines()
+            .find(|l| l.starts_with("VmHWM:"))
+            .expect("VmHWM");
+        let kb: u64 = line
+            .split_whitespace()
+            .nth(1)
+            .and_then(|v| v.parse().ok())
+            .expect("VmHWM value");
+        kb * 1024
+    }
+
+    /// Reset the kernel's peak-RSS watermark so the measurement covers only what follows.
+    /// Linux 4.0+; writing "5" to `clear_refs` resets `VmHWM` to the current RSS.
+    fn reset_peak_rss() {
+        let _ = std::fs::write("/proc/self/clear_refs", "5");
+    }
+
+    /// The pixel gate has to sit below what the axis limits allow, or it can never fire.
+    #[test]
+    fn the_oxipng_gate_is_reachable_within_the_decode_limits() {
+        const _: () = {
+            // imaging::decode_limits permits 12_000 x 12_000 = 144 MP. A gate above that would
+            // never skip anything.
+            assert!(CompressionWorker::OXIPNG_MAX_PIXELS < 12_000 * 12_000);
+            // ...and it must stay above a 48 MP camera, so real photos still get optimised.
+            assert!(CompressionWorker::OXIPNG_MAX_PIXELS >= 8_000_000);
+        };
+    }
+
+    /// The heavy-image gate has to classify the two cases the way the sizing assumed:
+    /// an ordinary phone photo must NOT serialise, and the giant must.
+    #[test]
+    fn the_heavy_gate_separates_a_phone_photo_from_a_giant() {
+        let dir = std::env::temp_dir().join(format!("es-heavy-{}", std::process::id()));
+        std::fs::create_dir_all(&dir).unwrap();
+
+        // 12 MP, the shape of a default phone capture.
+        let ordinary = dir.join("ordinary.jpg");
+        image::RgbImage::new(4032, 3024).save(&ordinary).unwrap();
+        let ordinary_peak = crate::services::imaging::estimated_processing_peak_bytes(
+            &ordinary,
+            CompressionWorker::DISPLAY_MAX_EDGE,
+        )
+        .expect("header readable");
+        assert!(
+            ordinary_peak <= crate::services::imaging::HEAVY_IMAGE_BYTES,
+            "a 12 MP photo estimated at {} MiB would serialise the common path",
+            ordinary_peak / 1048576
+        );
+
+        // The 64 MP RGBA case that measured ~516 MiB peak.
+        let giant = dir.join("giant.png");
+        image::RgbaImage::new(8000, 8000).save(&giant).unwrap();
+        let giant_peak = crate::services::imaging::estimated_processing_peak_bytes(
+            &giant,
+            CompressionWorker::DISPLAY_MAX_EDGE,
+        )
+        .expect("header readable");
+        assert!(
+            giant_peak > crate::services::imaging::HEAVY_IMAGE_BYTES,
+            "an 8000x8000 RGBA original estimated at only {} MiB would be allowed to run \
+             concurrently with another one — 2x its real ~516 MiB peak does not fit in 1 GiB",
+            giant_peak / 1048576
+        );
+        // The estimate must also be in the right ballpark, not merely on the right side of the
+        // threshold: 244 MiB decode + 262 MiB f32 resize intermediate.
+        assert!(
+            (400..700).contains(&(giant_peak / 1048576)),
+            "estimate {} MiB is far from the measured ~516 MiB peak",
+            giant_peak / 1048576
+        );
+
+        let _ = std::fs::remove_dir_all(&dir);
+    }
+
+    /// The OOM that took the container down, measured rather than argued.
+    ///
+    /// An 8000x8000 RGBA PNG passes admission: 256,000,000 bytes is just under the 256 MiB
+    /// `max_alloc`, and smooth content is a few MB on disk, far under any size cap. The old
+    /// code kept that ~244 MiB decode alive across an unbounded, multi-threaded oxipng run and
+    /// peaked at ~1250 MiB — inside a 1 GiB cgroup. Being SIGKILLed there is not a blip: the
+    /// row was already committed, so the boot backfill replayed the identical workload on every
+    /// restart.
+    ///
+    /// `#[ignore]` because it allocates ~250 MiB and takes a few seconds. Run explicitly:
+    ///   cargo test --release oom -- --ignored --nocapture --test-threads=1
+    /// It must run ALONE — `VmHWM` is per process, so a concurrent test would pollute it.
+    #[test]
+    #[ignore = "heavy: allocates ~250 MiB; run with --ignored --test-threads=1"]
+    fn a_large_png_stays_far_below_the_container_limit() {
+        const EDGE: u32 = 8_000;
+        let dir = std::env::temp_dir().join(format!("es-oom-{}", std::process::id()));
+        std::fs::create_dir_all(&dir).unwrap();
+        let original = dir.join("big.png");
+
+        // Smooth gradient: ~244 MiB decoded, a couple of MB on disk. That gap is the whole
+        // point — file size tells you nothing about what a PNG costs to process.
+        {
+            let mut buf = image::RgbaImage::new(EDGE, EDGE);
+            for (x, y, px) in buf.enumerate_pixels_mut() {
+                *px = image::Rgba([(x >> 5) as u8, (y >> 5) as u8, ((x + y) >> 6) as u8, 255]);
+            }
+            buf.save(&original).unwrap();
+        }
+
+        // Everything above is fixture setup, not the code under test.
+        reset_peak_rss();
+        let before = peak_rss_bytes();
+
+        write_image_derivatives(
+            Uuid::new_v4(),
+            &original,
+            "image/png",
+            &dir.join("preview.jpg"),
+            &dir.join("display.jpg"),
+        )
+        .expect("derivatives");
+
+        let peak = peak_rss_bytes();
+        let on_disk = std::fs::metadata(&original).unwrap().len();
+        eprintln!(
+            "input {:.2} MiB on disk ({EDGE}x{EDGE}); peak RSS {:.0} MiB (was {:.0} MiB before)",
+            on_disk as f64 / 1048576.0,
+            peak as f64 / 1048576.0,
+            before as f64 / 1048576.0
+        );
+
+        assert!(dir.join("preview.jpg").exists() && dir.join("display.jpg").exists());
+        // The container gets 1 GiB and runs two of these concurrently. 600 MiB is a generous
+        // ceiling that the old code (~1250 MiB) could not have met.
+        assert!(
+            peak < 600 * 1024 * 1024,
+            "peak RSS {} MiB — the decode is being held across oxipng again, or the pixel \
+             gate stopped firing",
+            peak / 1048576
+        );
+        let _ = std::fs::remove_dir_all(&dir);
     }
 }

backend/src/services/disk.rs

@@ -76,6 +76,17 @@ impl Default for DiskCache {
     }
 }
 
+/// UNCACHED free-space reading for the filesystem backing `path`.
+///
+/// Deliberately bypasses [`DiskCache`]. The cache exists for the quota poll, where a 15s-stale
+/// number is fine because it is only ever advisory. The export preflight is the opposite case: it
+/// decides whether to start writing a multi-GB archive, and the sibling export worker running
+/// concurrently can move free space by tens of gigabytes well inside the TTL. A stale reading there
+/// would authorise exactly the write that fills the disk.
+pub fn free_bytes(path: &Path) -> Option<u64> {
+    read_disk_for_path(path).map(|d| d.free)
+}
+
 /// Resolve the filesystem backing `media_path` and read its total/free bytes.
 ///
 /// Snapshots the mount table via `sysinfo`, then delegates the selection to the pure

backend/src/services/imaging.rs

@@ -0,0 +1,388 @@
+//! Shared image decoding.
+//!
+//! Exists so there is exactly ONE way to turn a file on disk into a `DynamicImage` in this
+//! codebase. Two properties have to hold everywhere an image is decoded, and both were
+//! previously re-derived per call site — which is how they drifted apart:
+//!
+//! - **EXIF orientation must be applied.** Phones do not rotate sensor data; they record how
+//!   the camera was held in a tag and store the pixels as shot. `image::open` and
+//!   `ImageReader::decode` both hand back the raw pixels and ignore that tag, and re-encoding
+//!   to JPEG writes no EXIF, so the derivative is permanently sideways while the untouched
+//!   original still renders upright. The compression worker was fixed; the export worker was
+//!   not, so every portrait photo came out sideways in the keepsake's HTML viewer.
+//! - **Decode limits must be set.** The upload body cap bounds the file on disk, but a small
+//!   file can decode to enormous dimensions (a ~1 MB image expanding to 50k×50k px), OOM-ing
+//!   the box. `image::open` applies NO limits at all, so the export path was also decoding
+//!   arbitrary user-supplied images unbounded.
+
+use anyhow::{Context, Result};
+use image::{DynamicImage, ImageDecoder};
+use std::path::Path;
+
+/// Bounds for any decode of user-supplied image data. The per-axis cap covers any real phone
+/// photo; `max_alloc` bounds the decoded buffer — but only because `decode_oriented` reserves
+/// against it explicitly, see there.
+///
+/// Sized against the deployment: the app container is capped at 1 GiB and the compression
+/// worker runs `compression_concurrency` decodes at once (default 2), so 256 MiB per decode
+/// leaves headroom for the resize buffers and the runtime.
+fn decode_limits() -> image::Limits {
+    let mut limits = image::Limits::default();
+    limits.max_image_width = Some(12_000);
+    limits.max_image_height = Some(12_000);
+    limits.max_alloc = Some(256 * 1024 * 1024);
+    limits
+}
+
+/// True when re-running the exact same work on the exact same bytes cannot possibly
+/// succeed, so retrying only burns wall-clock and log noise.
+///
+/// Deliberately narrow. Only the `ImageError` variants that are a property of the *input*
+/// count: the file will not shrink, gain codec support, or un-corrupt itself between
+/// attempts. `IoError` is excluded on purpose — EMFILE under load, or a momentarily
+/// unreadable file, is exactly the transient case the retry exists for. A FULL disk is the
+/// one io error that must not be retried either, but for a different reason and with a
+/// different remedy; see [`is_storage_full_error`].
+pub fn is_permanent_image_error(err: &anyhow::Error) -> bool {
+    err.chain().any(|cause| {
+        matches!(
+            cause.downcast_ref::<image::ImageError>(),
+            Some(
+                image::ImageError::Limits(_)
+                    | image::ImageError::Unsupported(_)
+                    | image::ImageError::Decoding(_)
+            )
+        )
+    })
+}
+
+/// True when the failure is the media filesystem being out of space.
+///
+/// Deliberately separate from [`is_permanent_image_error`], which is about the *input*. ENOSPC
+/// is about the *host*, and it is the one failure the retry loop actively makes worse: a disk
+/// does not drain during six seconds of backoff, so all three attempts fail identically while
+/// holding a compression permit that photos are queued behind.
+///
+/// The give-up path it fed was worse still. It refunded the guest's quota and soft-deleted the
+/// row while deliberately RETAINING the original — so the bytes stayed on the full disk, the
+/// photo vanished from the feed seconds after a `201 Created`, and the guest was handed back
+/// the quota to upload it again into the same full disk. Each round shrank free space further.
+pub fn is_storage_full_error(err: &anyhow::Error) -> bool {
+    fn is_full(io: &std::io::Error) -> bool {
+        // `StorageFull` is the portable classification; the raw ENOSPC catches the paths where
+        // the OS error was never mapped to a named kind.
+        io.kind() == std::io::ErrorKind::StorageFull || io.raw_os_error() == Some(28)
+    }
+    err.chain().any(|cause| {
+        // `image` wraps the io error in its own variant rather than exposing it as a source,
+        // so the plain downcast alone would miss every derivative-write failure.
+        cause.downcast_ref::<std::io::Error>().is_some_and(is_full)
+            || matches!(
+                cause.downcast_ref::<image::ImageError>(),
+                Some(image::ImageError::IoError(io)) if is_full(io)
+            )
+    })
+}
+
+/// Build a decoder for `path` with the budget enforced, WITHOUT reading any pixels.
+///
+/// Single source of truth for "may this image be decoded at all": both the upload
+/// admission check and the compression worker go through here, so they cannot disagree
+/// about what is acceptable.
+fn decoder_within_budget(path: &Path) -> Result<impl image::ImageDecoder> {
+    let mut reader = image::ImageReader::open(path)
+        .context("failed to open image")?
+        .with_guessed_format()
+        .context("failed to read image header")?;
+    let mut limits = decode_limits();
+    reader.limits(limits.clone());
+
+    // We need `into_decoder` rather than `decode()` to read the EXIF orientation tag before
+    // the pixels are consumed. But the two are NOT equivalent on safety: `decode()` performs
+    //
+    //     limits.reserve(decoder.total_bytes())?;
+    //
+    // between building the decoder and reading the image, and `into_decoder()` skips it (the
+    // crate's own FIXME concedes `from_decoder` doesn't compensate). Nothing else enforces
+    // `max_alloc` — the JPEG decoder's `set_limits` only checks support and dimensions — so
+    // without the line below the budget is inert and the ONLY bound is the per-axis cap. That
+    // leaves 12000x12000 decodable at 412 MiB, and two concurrent at 824 MiB against a 1 GiB
+    // container. Re-add it, exactly as `decode()` does.
+    let mut decoder = reader.into_decoder().context("failed to decode image")?;
+    limits
+        .reserve(decoder.total_bytes())
+        .context("image too large to decode within the memory budget")?;
+    decoder
+        .set_limits(limits)
+        .context("image too large to decode within the memory budget")?;
+    Ok(decoder)
+}
+
+/// Rough peak heap an image will cost to turn into derivatives, read from the HEADER only —
+/// no pixels are decoded. `None` when the header can't be read or the image is over budget
+/// (the caller is about to fail on it anyway).
+///
+/// Two terms, and the second is the one that surprises:
+///
+/// - the decoded buffer, `width * height * channels`; and
+/// - the resize intermediate. `image`'s Lanczos3 path accumulates in `f32`, so the buffer
+///   between the horizontal and vertical passes is `new_width * old_height * 4 channels * 4
+///   bytes` — 16 bytes per pixel-row-slot, not the 4 the output uses. For an 8000x8000
+///   original that is 262 MiB on top of a 244 MiB decode, measured. It is bigger than the
+///   decode for any tall image, which is why "the decode is bounded by max_alloc" was never
+///   the whole story.
+///
+/// Used to decide whether an image is heavy enough to need exclusive use of the box's memory
+/// headroom, NOT to reject anything.
+pub fn estimated_processing_peak_bytes(path: &Path, display_edge: u32) -> Option<u64> {
+    let decoder = decoder_within_budget(path).ok()?;
+    let (width, height) = decoder.dimensions();
+    let decoded = decoder.total_bytes();
+
+    // Aspect-preserving fit into `display_edge`, matching DynamicImage::resize. No downscale
+    // means no intermediate at all.
+    let intermediate = if width > display_edge || height > display_edge {
+        let ratio = f64::from(display_edge) / f64::from(width.max(height));
+        let new_width = (f64::from(width) * ratio).round().max(1.0) as u64;
+        new_width * u64::from(height) * 16
+    } else {
+        0
+    };
+
+    Some(decoded.saturating_add(intermediate))
+}
+
+/// Megapixels an image would decode to, or `None` if its header can't be read. Used only
+/// to put a concrete number in the message the guest sees.
+pub fn megapixels(path: &Path) -> Option<f64> {
+    let reader = image::ImageReader::open(path)
+        .ok()?
+        .with_guessed_format()
+        .ok()?;
+    let (w, h) = reader.into_dimensions().ok()?;
+    Some(f64::from(w) * f64::from(h) / 1_000_000.0)
+}
+
+/// True when an image cannot be decoded specifically because it would exceed the memory
+/// budget — read from the header, no pixels touched.
+///
+/// Called at upload admission so a guest who sends a 100 MP photo is told at the door, with
+/// a reason they can act on, instead of the upload being accepted with a 201 and then
+/// silently soft-deleted minutes later when the worker gives up on it.
+///
+/// Deliberately narrow: ONLY the budget. A corrupt, truncated or unsupported file also
+/// fails to build a decoder, but rejecting those here would change a contract the
+/// adversarial suite pins on purpose — acceptance follows the magic bytes, and a payload
+/// with a valid JPEG header is accepted regardless of what follows it. Those go to the
+/// compression worker as before, which handles them gracefully and (since the retry
+/// classifier) no longer burns backoff on them.
+pub fn exceeds_decode_budget(path: &Path) -> bool {
+    match decoder_within_budget(path) {
+        Ok(_) => false,
+        Err(e) => e.chain().any(|cause| {
+            matches!(
+                cause.downcast_ref::<image::ImageError>(),
+                Some(image::ImageError::Limits(_))
+            )
+        }),
+    }
+}
+
+/// Decode an image from disk with decompression-bomb limits applied and its EXIF
+/// orientation baked into the pixels.
+///
+/// Blocking — call inside `spawn_blocking`.
+pub fn decode_oriented(path: &Path) -> Result<DynamicImage> {
+    let mut decoder = decoder_within_budget(path)?;
+
+    // Cheap, and it happens BEFORE any pixels are read: an oversized image costs a header
+    // parse, not an allocation.
+    let orientation = decoder
+        .orientation()
+        .unwrap_or(image::metadata::Orientation::NoTransforms);
+    let mut img = DynamicImage::from_decoder(decoder).context("failed to decode image")?;
+    img.apply_orientation(orientation);
+    Ok(img)
+}
+
+/// Process-wide serialisation for memory-heavy image work.
+///
+/// The `app` container gets 1 GiB. A single 8000x8000 original measures ~516 MiB peak even with
+/// the decode correctly scoped, so two overlapping giants is an OOM kill — and the kernel kills
+/// the whole process, dropping every SSE stream and stranding every in-flight upload.
+///
+/// GLOBAL rather than a field on `CompressionWorker`, because the constraint is the container's
+/// memory and there is more than one producer of this work. The export's own image path
+/// (`services::export`) decodes and resizes every photo in the gallery — a thumbnail for each,
+/// plus a 2000px re-encode for every original over 5 MB — and it ran in a bare `spawn_blocking`
+/// with no permit at all. So "host taps Freigeben while the last phone photos are still
+/// compressing" put an export decode and a heavy compression job in the same cgroup at the same
+/// time, which is the scenario the permit exists to make impossible. Worse, it is self-repeating:
+/// the OOM kill marks the export failed, and `recover_exports` re-spawns it on boot into the same
+/// conditions.
+///
+/// Held across the blocking section and released on drop, including on error.
+pub static HEAVY_IMAGE_PERMITS: std::sync::LazyLock<tokio::sync::Semaphore> =
+    std::sync::LazyLock::new(|| tokio::sync::Semaphore::new(1));
+
+/// Estimated peak heap above which a job must take [`HEAVY_IMAGE_PERMITS`].
+///
+/// 150 MiB sits far above a normal phone photo (a 12 MP JPEG costs ~50 MiB all-in) so the common
+/// path never serialises, and far below the point where two jobs stop fitting in the container.
+pub const HEAVY_IMAGE_BYTES: u64 = 150 * 1024 * 1024;
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Shared with the e2e suite rather than duplicating 568 KiB of binary: the same file
+    /// drives `02-upload/oversized-image` so both layers assert on one artefact.
+    const HUGE: &str = concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/../e2e/fixtures/media/huge-99mp.jpg"
+    );
+
+    #[test]
+    fn rejects_an_image_that_would_blow_the_allocation_budget() {
+        // 11000x9000 = 99 MP. Deliberately UNDER the 12000px per-axis cap, so the axis check
+        // cannot reject it — the allocation budget is the only thing that can, which is
+        // exactly what makes this a regression test rather than a restatement of the axis cap.
+        // 283 MiB decoded as RGB8 against a 256 MiB budget, from 568 KiB on disk.
+        //
+        // This failed before the guard was restored: `ImageReader::decode` performs
+        // `limits.reserve(decoder.total_bytes())`, and `into_decoder()` — which we need for
+        // the EXIF tag — skips it, so `max_alloc` was inert and this decoded happily.
+        // Map the Ok arm to its dimensions first: on failure `expect_err` Debug-prints the
+        // value, and Debug on a DynamicImage dumps every pixel — 283 MiB of output.
+        let err = decode_oriented(Path::new(HUGE))
+            .map(|img| (img.width(), img.height()))
+            .expect_err("a 99 MP image must be refused, not allocated");
+        let msg = format!("{err:#}");
+        assert!(
+            msg.to_lowercase().contains("limit") || msg.to_lowercase().contains("memory"),
+            "expected a limits error, got: {msg}"
+        );
+    }
+
+    #[test]
+    fn an_oversized_image_is_a_permanent_failure() {
+        // The retry loop must not burn 2s + 4s of backoff on this: the file will not shrink
+        // between attempts, so all three attempts reach the identical conclusion.
+        let err = decode_oriented(Path::new(HUGE))
+            .map(|img| (img.width(), img.height()))
+            .expect_err("fixture must exceed the budget");
+        assert!(
+            is_permanent_image_error(&err),
+            "a Limits error can never succeed on retry: {err:#}"
+        );
+    }
+
+    #[test]
+    fn a_plain_io_error_is_not_permanent() {
+        // The mirror that keeps the classifier honest. EMFILE under load, or a momentary
+        // unreadable file, is exactly what the retry exists for — misclassifying those as
+        // permanent would turn a transient blip back into the data loss round 1 fixed.
+        // (A FULL disk is its own case now; see the storage-full tests below.)
+        let err = decode_oriented(Path::new("/nonexistent/definitely-not-here.jpg"))
+            .map(|img| (img.width(), img.height()))
+            .expect_err("a missing file must error");
+        assert!(
+            !is_permanent_image_error(&err),
+            "an IO error must stay retryable: {err:#}"
+        );
+    }
+
+    #[test]
+    fn a_full_disk_is_recognised_through_both_wrappers() {
+        // The two shapes ENOSPC actually arrives in. A bare io::Error is what `tokio::fs` and
+        // `std::fs` produce; the `image` crate wraps its own in `ImageError::IoError`, which is
+        // NOT reachable via `source()` — so a chain walk that only downcast to io::Error would
+        // miss every derivative-write failure, i.e. the exact case this classifier exists for.
+        let bare = anyhow::Error::from(std::io::Error::from(std::io::ErrorKind::StorageFull))
+            .context("failed to write the preview");
+        assert!(is_storage_full_error(&bare), "bare io::Error: {bare:#}");
+
+        let wrapped = anyhow::Error::from(image::ImageError::IoError(std::io::Error::from(
+            std::io::ErrorKind::StorageFull,
+        )))
+        .context("failed to save the display derivative");
+        assert!(
+            is_storage_full_error(&wrapped),
+            "ImageError::IoError: {wrapped:#}"
+        );
+    }
+
+    #[test]
+    fn an_ordinary_io_error_is_not_a_full_disk() {
+        // Keeps the classifier from swallowing the general case: only ENOSPC may skip the retry
+        // and take the keep-the-row branch. Anything else must still be retried and, if it keeps
+        // failing, soft-deleted as before.
+        let missing = decode_oriented(Path::new("/nonexistent/definitely-not-here.jpg"))
+            .map(|img| (img.width(), img.height()))
+            .expect_err("a missing file must error");
+        assert!(
+            !is_storage_full_error(&missing),
+            "a missing file is not a full disk: {missing:#}"
+        );
+    }
+
+    #[test]
+    fn admission_rejects_only_the_over_budget_case() {
+        // Admission and processing must agree about SIZE — a photo accepted at the door and
+        // then rejected by the worker for being too big is the failure this pair prevents.
+        assert!(
+            exceeds_decode_budget(Path::new(HUGE)),
+            "admission must reject what the decoder rejects for size"
+        );
+        let ordinary = concat!(
+            env!("CARGO_MANIFEST_DIR"),
+            "/../e2e/fixtures/media/portrait-exif6.jpg"
+        );
+        assert!(
+            !exceeds_decode_budget(Path::new(ordinary)),
+            "admission must accept an ordinary photo"
+        );
+    }
+
+    #[test]
+    fn admission_does_not_reject_a_merely_undecodable_file() {
+        // The narrowing that keeps the adversarial contract intact: a payload with valid
+        // JPEG magic bytes and nothing behind them cannot be decoded, but acceptance follows
+        // the magic bytes by design (07-adversarial/file-upload-attacks). It is the worker's
+        // job to fail it, not admission's — admission is only the resource guard.
+        let dir = std::env::temp_dir().join("eventsnap-imaging-test");
+        std::fs::create_dir_all(&dir).expect("tmp dir");
+        let stub = dir.join("magic-only.jpg");
+        let mut bytes = vec![0u8; 1024];
+        bytes[..3].copy_from_slice(&[0xFF, 0xD8, 0xFF]);
+        std::fs::write(&stub, &bytes).expect("write stub");
+
+        assert!(
+            !exceeds_decode_budget(&stub),
+            "a corrupt file is not an over-budget file"
+        );
+        assert!(
+            decode_oriented(&stub)
+                .map(|i| (i.width(), i.height()))
+                .is_err(),
+            "...but it must still fail in the worker"
+        );
+        let _ = std::fs::remove_file(&stub);
+    }
+
+    #[test]
+    fn still_decodes_an_ordinary_photo_and_applies_orientation() {
+        // The guard must not have become a blanket refusal. This fixture is 40x20 stored with
+        // EXIF Orientation=6, so a correct decode returns it rotated to 20x40 portrait.
+        let path = concat!(
+            env!("CARGO_MANIFEST_DIR"),
+            "/../e2e/fixtures/media/portrait-exif6.jpg"
+        );
+        let img = decode_oriented(Path::new(path)).expect("an ordinary photo must decode");
+        assert_eq!(
+            (img.width(), img.height()),
+            (20, 40),
+            "EXIF orientation must still be applied after restoring the guard"
+        );
+    }
+}

backend/src/services/maintenance.rs

@@ -9,10 +9,13 @@
 //!    users staring at a spinner. Resetting them on startup recovers gracefully.
 //!
 //! 2. **Periodic tasks** — pruning that should happen "every hour" rather than per
-//!    request: expired sessions (otherwise the table grows unboundedly), and the
+//!    request: expired sessions (otherwise the table grows unboundedly), the
 //!    rate-limiter's in-memory windows (so keys for IPs that left long ago don't
-//!    accumulate).
+//!    accumulate), and the media of soft-deleted uploads — both the ones whose compression
+//!    permanently failed and the ones a guest or host deliberately removed — which are
+//!    retained for a recovery window and then reclaimed.
 
+use std::path::PathBuf;
 use std::time::Duration;
 
 use sqlx::PgPool;
@@ -20,6 +23,42 @@ use sqlx::PgPool;
 use crate::services::rate_limiter::RateLimiter;
 use crate::services::sse_tickets::SseTicketStore;
 
+/// How long a permanently-failed upload's original is kept on disk before it is
+/// reclaimed.
+///
+/// The compression worker stops deleting originals on failure — a transient error must
+/// never destroy the guest's only copy of a photo they can't retake. But the row is
+/// soft-deleted and the uploader's quota IS refunded, so without a sweep those bytes are
+/// invisible, unowned, and free: a reproducible codec failure lets one guest accumulate
+/// orphans at no personal cost, and because `active_uploaders` counts only users with
+/// non-deleted uploads, dropping out of that count actually RAISES everyone's per-user
+/// ceiling while the disk gets fuller.
+///
+/// Two weeks is comfortably longer than any single event, so an operator investigating a
+/// failed upload still has the file, while the leak stays bounded.
+const FAILED_ORIGINAL_RETENTION_DAYS: i64 = 14;
+
+/// How long a DELIBERATELY deleted upload's files are kept before they are reclaimed.
+///
+/// The same leak, reached by the ordinary path rather than the exceptional one.
+/// `soft_delete_in_event` stamps `deleted_at` and refunds `total_upload_bytes`, but nothing ever
+/// removed the bytes — so the quota stopped bounding the disk. Upload 500 MB, delete, quota is back
+/// to zero, upload another 500 MB: not an attack, just a guest curating their camera roll, which is
+/// what people do. The host then sees guests hitting "Du hast dein Upload-Limit erreicht" while the
+/// admin widget shows a disk full of files no upload row points at, and the quota message is
+/// actively misleading because the space really is gone — just not to anyone the accounting can
+/// name.
+///
+/// Much shorter than the failure window on purpose. Fourteen days outlives the whole event, so a
+/// deliberate delete would never reclaim anything while it mattered. A day still gives an operator
+/// a recovery window for a mis-tap.
+///
+/// NOTE what this does NOT do: within the window the bytes are still spent and still unaccounted,
+/// so a guest deleting and re-uploading through an eight-hour event can outrun the sweep. Bounding
+/// that would mean holding the quota until the file is actually reclaimed rather than refunding at
+/// `deleted_at` — a deliberate trade, and the reason the low-disk warning exists.
+const DELETED_UPLOAD_RETENTION_HOURS: i64 = 24;
+
 /// Reset rows left in flight by a previous crashed instance. Run once on startup,
 /// before the HTTP server starts taking requests, so users never observe the
 /// half-state.
@@ -81,26 +120,272 @@ pub async fn startup_recovery(pool: &PgPool) {
     }
 }
 
+/// How long a file in `originals/` may exist without a database row before it is treated as
+/// abandoned.
+///
+/// This window is the ONLY thing making the sweep safe, because the upload handler renames the
+/// temp file into its final path BEFORE committing the row: for a short moment a perfectly
+/// healthy upload legitimately looks exactly like an orphan. Six hours is far beyond any live
+/// request (a 576 MiB body over a bad venue uplink is minutes, and the request itself is bounded
+/// by the reverse proxy) while still reclaiming the leak inside a single event.
+///
+/// DO NOT SHORTEN THIS to make a test faster — a value below the longest possible in-flight
+/// upload deletes photos out from under the request that is committing them.
+const ORPHAN_UPLOAD_RETENTION_HOURS: u64 = 6;
+
 /// Spawns a background task that periodically:
 /// - deletes session rows whose `expires_at` is more than a day in the past
 /// - prunes the in-memory rate-limiter HashMap of empty windows
 /// - drops expired SSE tickets (30s TTL but the map keeps the slot until pruned)
 ///
 /// Cadence is 1h — fine for both jobs at our scale.
-pub fn spawn_periodic_tasks(pool: PgPool, rate_limiter: RateLimiter, sse_tickets: SseTicketStore) {
+pub fn spawn_periodic_tasks(
+    pool: PgPool,
+    rate_limiter: RateLimiter,
+    sse_tickets: SseTicketStore,
+    media_path: PathBuf,
+) {
+    // Supervised, because this one task carries EVERY piece of recurring hygiene in the app:
+    // session pruning, media reclamation, the orphan-temp sweep, and the rate-limiter and
+    // SSE-ticket maps. As a bare `tokio::spawn` with no retained handle, a single panic anywhere
+    // inside it stopped all five permanently and silently — no log line, no symptom until the
+    // disk or a HashMap grew into one. The supervisor re-spawns and, just as importantly, says
+    // so; it can never spin hot because the inner loop only returns by dying.
     tokio::spawn(async move {
-        let mut tick = tokio::time::interval(Duration::from_secs(3600));
-        // Fire the first tick immediately, then hourly.
-        tick.tick().await;
         loop {
-            tick.tick().await;
-            cleanup_sessions(&pool).await;
-            rate_limiter.prune();
-            sse_tickets.prune();
+            let inner = tokio::spawn(periodic_loop(
+                pool.clone(),
+                rate_limiter.clone(),
+                sse_tickets.clone(),
+                media_path.clone(),
+            ));
+            match inner.await {
+                Ok(()) => tracing::error!("periodic maintenance loop returned; restarting it"),
+                Err(e) => {
+                    tracing::error!(error = ?e, "periodic maintenance task died; restarting it")
+                }
+            }
+            tokio::time::sleep(Duration::from_secs(60)).await;
         }
     });
 }
 
+/// The actual hygiene loop. Never returns in normal operation — see the supervisor above.
+async fn periodic_loop(
+    pool: PgPool,
+    rate_limiter: RateLimiter,
+    sse_tickets: SseTicketStore,
+    media_path: PathBuf,
+) {
+    // A crash left whatever the previous process was mid-upload behind, and the first periodic
+    // tick is an hour away — sweep once up front so a restart is also a cleanup.
+    sweep_orphan_upload_temps(&media_path).await;
+
+    let mut tick = tokio::time::interval(Duration::from_secs(3600));
+    // Fire the first tick immediately, then hourly.
+    tick.tick().await;
+    loop {
+        tick.tick().await;
+        cleanup_sessions(&pool).await;
+        cleanup_deleted_media(&pool, &media_path).await;
+        sweep_orphan_upload_temps(&media_path).await;
+        // Runs AFTER the .tmp sweep, and covers the class that one structurally cannot see:
+        // an original that was renamed to its final name but whose transaction never
+        // committed. Those have no row, so `cleanup_deleted_media` (row-driven) can never
+        // find them, and `sweep_orphan_upload_temps` skips them because they no longer end
+        // in `.tmp` — they were permanently unowned, silently shrinking the free disk that
+        // `compute_storage_quota` divides among guests.
+        sweep_orphan_originals(&pool, &media_path).await;
+        rate_limiter.prune();
+        sse_tickets.prune();
+    }
+}
+
+/// How long an upload's `.tmp` file must have been untouched before it is treated as abandoned.
+///
+/// This is an age on the MODIFICATION time, not on creation, and that is what makes an hour
+/// safe rather than reckless: a live upload is being written to continuously, so its mtime keeps
+/// advancing and it can never age into the sweep no matter how slow the connection. The clock
+/// only starts once the writer stops — i.e. once the upload is genuinely dead.
+const ORPHAN_TEMP_MAX_AGE: Duration = Duration::from_secs(3600);
+
+/// Reclaim `.tmp` files left in the media tree by uploads that never finished.
+///
+/// `stream_field_to_file` removes its temp file on every error return, which covers everything
+/// the handler can see. It cannot cover the case that actually happens at a party: the client
+/// simply goes away — a phone sleeps, a guest walks out of range, the PWA is evicted mid-video —
+/// and axum DROPS the handler future rather than returning an error, so no cleanup code runs at
+/// all. The shutdown backstop force-exits in-flight handlers for the same net effect.
+///
+/// Nothing else reclaims these. `cleanup_deleted_media` only visits rows with `deleted_at`, and
+/// an abandoned upload never got a row; `export::sweep_orphan_temps` is only ever pointed at the
+/// exports volume. So before this, every abandonment stranded up to `max_video_size_mb` of
+/// unowned bytes permanently — and worse than merely leaking, they were subtracted from what
+/// everyone else could upload, because the per-user quota is computed from live free disk
+/// (`compute_storage_quota`). On a 40 GB disk shared with `postgres_data`, an evening of flaky
+/// venue wifi could take the event down.
+async fn sweep_orphan_upload_temps(media_path: &std::path::Path) {
+    let originals = media_path.join("originals");
+    let mut event_dirs = match tokio::fs::read_dir(&originals).await {
+        Ok(d) => d,
+        // Absent before the first upload — not a problem worth logging every hour.
+        Err(e) if e.kind() == std::io::ErrorKind::NotFound => return,
+        Err(e) => {
+            tracing::warn!(error = ?e, path = %originals.display(), "orphan temp sweep: unreadable");
+            return;
+        }
+    };
+
+    let mut reclaimed = 0usize;
+    let mut bytes = 0u64;
+    while let Ok(Some(event_dir)) = event_dirs.next_entry().await {
+        let Ok(mut files) = tokio::fs::read_dir(event_dir.path()).await else {
+            continue;
+        };
+        while let Ok(Some(file)) = files.next_entry().await {
+            let path = file.path();
+            if path.extension().is_none_or(|e| e != "tmp") {
+                continue;
+            }
+            let Ok(meta) = file.metadata().await else {
+                continue;
+            };
+            // No mtime (or a clock that moved backwards) means we cannot show the file is
+            // abandoned, and deleting a live upload is far worse than leaking one temp file.
+            let abandoned = meta
+                .modified()
+                .ok()
+                .and_then(|m| m.elapsed().ok())
+                .is_some_and(|age| age >= ORPHAN_TEMP_MAX_AGE);
+            if !abandoned {
+                continue;
+            }
+            match tokio::fs::remove_file(&path).await {
+                Ok(()) => {
+                    reclaimed += 1;
+                    bytes += meta.len();
+                }
+                Err(e) if e.kind() == std::io::ErrorKind::NotFound => {}
+                Err(e) => {
+                    tracing::warn!(error = ?e, path = %path.display(),
+                        "orphan temp sweep: could not reclaim")
+                }
+            }
+        }
+    }
+
+    if reclaimed > 0 {
+        tracing::info!(
+            "orphan temp sweep: reclaimed {reclaimed} abandoned upload temp file(s), {} MiB",
+            bytes / (1024 * 1024)
+        );
+    }
+}
+
+/// Reclaim the media of soft-deleted uploads once they are past their retention window.
+///
+/// ONLY ever touches rows with `deleted_at IS NOT NULL`, so it can never reach a live upload. Two
+/// classes, two windows, because the two deletes mean different things:
+///
+/// - a compression failure the guest didn't ask for and may want investigated —
+///   [`FAILED_ORIGINAL_RETENTION_DAYS`];
+/// - a deliberate removal by the guest or the host — [`DELETED_UPLOAD_RETENTION_HOURS`].
+///
+/// ALL FOUR paths are reclaimed, not just the original. The previous version cleared
+/// `original_path` alone, which was right for its only case (a failed compression produces no
+/// derivatives) but wrong the moment the sweep reaches a successfully processed upload: preview,
+/// display and thumbnail are each a separate file on disk, none of them counted in
+/// `original_size_bytes`, and nothing else ever removed them.
+///
+/// Every column is cleared in the same pass, which makes the sweep idempotent and stops a later run
+/// re-reporting files that are already gone. The ROW is kept: it is the audit trail, it costs a few
+/// hundred bytes, and `backfill_stale_derivatives` is guarded on `deleted_at IS NULL` so a nulled
+/// `preview_path` can never make it regenerate what was just reclaimed.
+async fn cleanup_deleted_media(pool: &PgPool, media_path: &std::path::Path) {
+    type Row = (
+        uuid::Uuid,
+        String,
+        Option<String>,
+        Option<String>,
+        Option<String>,
+    );
+    let rows = sqlx::query_as::<_, Row>(
+        "SELECT id, original_path, preview_path, display_path, thumbnail_path FROM upload
+         WHERE deleted_at IS NOT NULL
+           AND CASE WHEN compression_status = 'failed'
+                    THEN deleted_at < NOW() - ($1 || ' days')::interval
+                    ELSE deleted_at < NOW() - ($2 || ' hours')::interval
+               END
+           AND (original_path <> '' OR preview_path IS NOT NULL
+                OR display_path IS NOT NULL OR thumbnail_path IS NOT NULL)",
+    )
+    .bind(FAILED_ORIGINAL_RETENTION_DAYS.to_string())
+    .bind(DELETED_UPLOAD_RETENTION_HOURS.to_string())
+    .fetch_all(pool)
+    .await;
+
+    let rows = match rows {
+        Ok(r) => r,
+        Err(e) => {
+            tracing::warn!(error = ?e, "deleted-media sweep query failed");
+            return;
+        }
+    };
+    if rows.is_empty() {
+        return;
+    }
+
+    let mut reclaimed = 0usize;
+    for (id, original, preview, display, thumbnail) in rows {
+        let paths: Vec<String> = std::iter::once(original)
+            .filter(|p| !p.is_empty())
+            .chain([preview, display, thumbnail].into_iter().flatten())
+            .collect();
+
+        // All-or-nothing per row: the columns are only cleared once every file for that upload is
+        // gone. Clearing after a partial success would strand the survivors with nothing pointing
+        // at them — the same unowned-bytes state this sweep exists to drain.
+        let mut all_gone = true;
+        for rel in &paths {
+            let absolute = media_path.join(rel);
+            match tokio::fs::remove_file(&absolute).await {
+                Ok(()) => reclaimed += 1,
+                // Already gone (manual cleanup, restored backup) — still counts as reclaimed for
+                // the purpose of clearing the columns, or the row is re-selected every hour forever.
+                Err(e) if e.kind() == std::io::ErrorKind::NotFound => {}
+                Err(e) => {
+                    tracing::warn!(error = ?e, %id, path = %absolute.display(),
+                        "could not reclaim deleted media; leaving the row for the next sweep");
+                    all_gone = false;
+                }
+            }
+        }
+        if !all_gone {
+            continue;
+        }
+
+        if let Err(e) = sqlx::query(
+            "UPDATE upload SET original_path = '', preview_path = NULL,
+                    display_path = NULL, thumbnail_path = NULL
+             WHERE id = $1",
+        )
+        .bind(id)
+        .execute(pool)
+        .await
+        {
+            tracing::warn!(error = ?e, %id, "reclaimed the files but could not clear the paths");
+        }
+    }
+
+    if reclaimed > 0 {
+        tracing::info!(
+            "reclaimed {reclaimed} file(s) from soft-deleted uploads (deliberate deletes after \
+             {DELETED_UPLOAD_RETENTION_HOURS}h, compression failures after \
+             {FAILED_ORIGINAL_RETENTION_DAYS}d)"
+        );
+    }
+}
+
 async fn cleanup_sessions(pool: &PgPool) {
     match sqlx::query("DELETE FROM session WHERE expires_at < NOW() - INTERVAL '1 day'")
         .execute(pool)
@@ -113,3 +398,192 @@ async fn cleanup_sessions(pool: &PgPool) {
         Err(e) => tracing::warn!("session cleanup failed: {e:#}"),
     }
 }
+
+/// Reclaim files in `originals/` that no upload row references.
+///
+/// The backstop behind [`TempFileGuard`](crate::handlers::upload). The guard covers the
+/// process that is running; this covers the process that was killed — a SIGKILL, an OOM, or a
+/// power cut leaves whatever bytes had been written with no `Drop` to reclaim them, and those
+/// files are then permanently invisible: they have no row, so `cleanup_deleted_media` (which is
+/// row-driven) can never see them, and they are not counted against any quota while still
+/// consuming the free disk that `compute_storage_quota` divides among guests. On a single box
+/// where all three volumes share a filesystem, that ends with Postgres unable to write WAL.
+///
+/// Two classes:
+/// - `*.tmp` — an upload that never got as far as being renamed. Always safe past the window.
+/// - everything else — a final-named original whose commit never happened.
+async fn sweep_orphan_originals(pool: &PgPool, media_path: &std::path::Path) {
+    let originals = media_path.join("originals");
+    let cutoff = Duration::from_secs(ORPHAN_UPLOAD_RETENTION_HOURS * 3600);
+
+    // originals/{event_slug}/{uuid}.{ext} — one level of per-event directories.
+    let mut event_dirs = match tokio::fs::read_dir(&originals).await {
+        Ok(rd) => rd,
+        // Nothing uploaded yet; the directory is created lazily by the upload handler.
+        Err(_) => return,
+    };
+
+    let mut candidates: Vec<(String, std::path::PathBuf)> = Vec::new();
+    let mut temps_removed = 0u32;
+
+    while let Ok(Some(event_dir)) = event_dirs.next_entry().await {
+        if !event_dir
+            .file_type()
+            .await
+            .map(|t| t.is_dir())
+            .unwrap_or(false)
+        {
+            continue;
+        }
+        let slug = event_dir.file_name().to_string_lossy().to_string();
+        let Ok(mut files) = tokio::fs::read_dir(event_dir.path()).await else {
+            continue;
+        };
+        while let Ok(Some(entry)) = files.next_entry().await {
+            let Ok(meta) = entry.metadata().await else {
+                continue;
+            };
+            if !meta.is_file() {
+                continue;
+            }
+            // Too young to judge: an upload committing RIGHT NOW is indistinguishable from an
+            // orphan, because the rename precedes the commit.
+            let recent = meta
+                .modified()
+                .ok()
+                .and_then(|m| m.elapsed().ok())
+                .is_none_or(|age| age < cutoff);
+            if recent {
+                continue;
+            }
+
+            let name = entry.file_name().to_string_lossy().to_string();
+            if name.ends_with(".tmp") {
+                // A `.tmp` never has a row by construction — no DB check needed.
+                if tokio::fs::remove_file(entry.path()).await.is_ok() {
+                    temps_removed += 1;
+                }
+                continue;
+            }
+            candidates.push((format!("originals/{slug}/{name}"), entry.path()));
+        }
+    }
+
+    if temps_removed > 0 {
+        tracing::warn!(
+            "reclaimed {temps_removed} abandoned upload temp file(s) older than \
+             {ORPHAN_UPLOAD_RETENTION_HOURS}h"
+        );
+    }
+    if candidates.is_empty() {
+        return;
+    }
+
+    // One query per batch, not one per file: a backlog of thousands of orphans must not turn
+    // into thousands of round trips on an hourly timer.
+    let mut orphans_removed = 0u32;
+    for chunk in candidates.chunks(500) {
+        let paths: Vec<String> = chunk.iter().map(|(rel, _)| rel.clone()).collect();
+        // NO `deleted_at IS NULL` FILTER HERE. A soft-deleted row still points at its file
+        // during its retention window, and reclaiming that file is `cleanup_deleted_media`'s
+        // job — filtering here would race the two sweeps and destroy the exact files the
+        // recovery window exists to preserve.
+        let unreferenced: Result<Vec<(String,)>, _> = sqlx::query_as(
+            "SELECT p FROM unnest($1::text[]) AS p
+              WHERE NOT EXISTS (SELECT 1 FROM upload u WHERE u.original_path = p)",
+        )
+        .bind(&paths)
+        .fetch_all(pool)
+        .await;
+        let unreferenced = match unreferenced {
+            Ok(rows) => rows,
+            Err(e) => {
+                tracing::warn!(error = ?e, "orphan-original sweep query failed");
+                return;
+            }
+        };
+        for (rel,) in unreferenced {
+            if let Some((_, abs)) = chunk.iter().find(|(r, _)| *r == rel)
+                && tokio::fs::remove_file(abs).await.is_ok()
+            {
+                orphans_removed += 1;
+            }
+        }
+    }
+
+    if orphans_removed > 0 {
+        tracing::warn!(
+            "reclaimed {orphans_removed} original(s) with no upload row, older than \
+             {ORPHAN_UPLOAD_RETENTION_HOURS}h"
+        );
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Build `<root>/originals/<event>/<name>` with `len` bytes, optionally back-dating its mtime
+    /// by `age`. Back-dating is the only way to test the sweep without sleeping through an hour.
+    fn temp_file(root: &std::path::Path, name: &str, len: usize, age: Option<Duration>) {
+        let dir = root.join("originals").join("wedding");
+        std::fs::create_dir_all(&dir).expect("create dir");
+        let path = dir.join(name);
+        let f = std::fs::File::create(&path).expect("create file");
+        std::io::Write::write_all(&mut &f, &vec![0u8; len]).expect("write");
+        if let Some(age) = age {
+            let when = std::time::SystemTime::now() - age;
+            f.set_modified(when).expect("set mtime");
+        }
+    }
+
+    fn exists(root: &std::path::Path, name: &str) -> bool {
+        root.join("originals").join("wedding").join(name).exists()
+    }
+
+    /// The two halves of the guarantee in one pass: an abandoned temp is reclaimed, and a temp
+    /// that is still being written to is NOT — the second matters more, because deleting a live
+    /// upload's temp file would corrupt a photo that was about to succeed.
+    #[tokio::test]
+    async fn the_sweep_reclaims_abandoned_temps_and_spares_live_ones() {
+        let root = std::env::temp_dir().join(format!("es-sweep-{}", uuid::Uuid::new_v4()));
+
+        // Abandoned: the writer died over an hour ago and nothing has touched it since.
+        temp_file(
+            &root,
+            "dead.tmp",
+            2048,
+            Some(ORPHAN_TEMP_MAX_AGE + Duration::from_secs(60)),
+        );
+        // Live: an upload in progress keeps advancing its mtime, so it always looks young —
+        // this is why the threshold is on modification time and not on creation time.
+        temp_file(&root, "inflight.tmp", 2048, None);
+        // A committed original. The sweep must only ever consider `.tmp`.
+        temp_file(&root, "keeper.jpg", 2048, Some(Duration::from_secs(86_400)));
+
+        sweep_orphan_upload_temps(&root).await;
+
+        assert!(
+            !exists(&root, "dead.tmp"),
+            "an abandoned temp must be reclaimed"
+        );
+        assert!(
+            exists(&root, "inflight.tmp"),
+            "a temp still being written to must survive — deleting it destroys a live upload"
+        );
+        assert!(
+            exists(&root, "keeper.jpg"),
+            "the sweep must never touch a committed original"
+        );
+
+        let _ = std::fs::remove_dir_all(&root);
+    }
+
+    /// Runs on every boot and every hour, so a media tree that does not exist yet (before the
+    /// first upload) must be a silent no-op rather than an error logged 24 times a day.
+    #[tokio::test]
+    async fn a_missing_media_tree_is_not_an_error() {
+        let root = std::env::temp_dir().join(format!("es-sweep-absent-{}", uuid::Uuid::new_v4()));
+        sweep_orphan_upload_temps(&root).await; // must simply return
+    }
+}

backend/src/services/media_total.rs

@@ -0,0 +1,104 @@
+//! Cached sum of all media bytes the event is holding.
+//!
+//! The upload gate needs to know "how big would the keepsake be if we accept this file", because
+//! the archive needs room for BOTH halves at once (`export::required_free_bytes` is
+//! `media × 1.1 × 2` — the ZIP and the HTML viewer are each gallery-sized). Asking that question
+//! per upload has to be cheap, and it has to be cheap on the busiest write path in the app.
+//!
+//! `export::estimate_export_bytes` answers the same question exactly, but it aggregates
+//! `original_size_bytes` across every upload row joined to `user` — fine once per release,
+//! wasteful per upload and growing all evening. This sums `user.total_upload_bytes` instead:
+//! one row per guest (~100), already maintained transactionally by the quota path, already
+//! refunded on delete.
+//!
+//! The two differ slightly — this one counts uploads belonging to banned or hidden users, which
+//! the export filters out. That skew is in the SAFE direction: it over-estimates the archive, so
+//! the gate closes marginally early rather than marginally late. Never swap it for a cheaper
+//! query that could under-estimate; an under-estimate authorises the very upload that makes the
+//! keepsake unbuildable, which is the failure this exists to prevent.
+
+use std::sync::{Arc, RwLock};
+use std::time::{Duration, Instant};
+
+use sqlx::PgPool;
+
+/// How long a reading is trusted. Shorter than [`crate::services::disk`]'s TTL because this
+/// number only ever grows and does so in the same request path that reads it — a stale value
+/// under-counts the newest uploads, and under-counting is the direction that matters.
+const TTL: Duration = Duration::from_secs(5);
+
+/// Cheap-to-clone cache of the event's total media bytes. Lives in `AppState`.
+#[derive(Clone)]
+pub struct MediaTotalCache {
+    inner: Arc<RwLock<Option<(i64, Instant)>>>,
+}
+
+impl MediaTotalCache {
+    pub fn new() -> Self {
+        Self {
+            inner: Arc::new(RwLock::new(None)),
+        }
+    }
+
+    /// Drop the cached reading so the next `get()` re-queries.
+    ///
+    /// Used by the e2e TRUNCATE endpoint for the same reason `DiskCache::invalidate` exists:
+    /// truncation removes every upload, and a surviving reading would make the next test's
+    /// gate compute against the previous test's data.
+    pub fn invalidate(&self) {
+        *self.inner.write().unwrap() = None;
+    }
+
+    /// Total bytes of media the event is holding, cached for [`TTL`].
+    ///
+    /// Returns 0 when the query fails. That is a deliberate FAIL-OPEN, consistent with the
+    /// quota path and the export preflight: a database blip must not turn into "every upload
+    /// refused". The disk-space half of the gate still applies, so a failure here degrades the
+    /// check to the old flat-reserve behaviour rather than disabling it.
+    pub async fn get(&self, pool: &PgPool) -> i64 {
+        if let Some((bytes, at)) = *self.inner.read().unwrap()
+            && at.elapsed() < TTL
+        {
+            return bytes;
+        }
+        let queried = sqlx::query_scalar::<_, Option<i64>>(
+            "SELECT SUM(total_upload_bytes)::bigint FROM \"user\"",
+        )
+        .fetch_one(pool)
+        .await;
+
+        let bytes = match queried {
+            Ok(v) => v.unwrap_or(0).max(0),
+            Err(e) => {
+                // FAIL OPEN, but do NOT cache the failure, and do NOT let it pass silently.
+                //
+                // Storing 0 here pinned the gate's view of the event at "empty" for the whole
+                // TTL. During that window `media_after` is just this upload, `keepsake_needs`
+                // collapses to ~2.2x one file, and the gate degrades to the flat 10 GB reserve —
+                // precisely the behaviour the two-halves design replaced, reappearing with no
+                // trace in the log. And the trigger correlates with the danger: with
+                // `max_connections = 10` and a 5s acquire timeout, this query fails exactly when
+                // a burst is in progress.
+                //
+                // Falling back to the LAST GOOD reading (however stale) is strictly better than
+                // 0: the total only ever grows, so a stale value under-counts slightly, while 0
+                // under-counts by everything.
+                let previous = self.inner.read().unwrap().map(|(b, _)| b);
+                tracing::warn!(
+                    error = %e,
+                    fallback_bytes = previous.unwrap_or(0),
+                    "media total query failed; upload gate is running on a stale reading"
+                );
+                return previous.unwrap_or(0);
+            }
+        };
+        *self.inner.write().unwrap() = Some((bytes, Instant::now()));
+        bytes
+    }
+}
+
+impl Default for MediaTotalCache {
+    fn default() -> Self {
+        Self::new()
+    }
+}

backend/src/services/mod.rs

@@ -2,6 +2,10 @@ pub mod compression;
 pub mod config;
 pub mod disk;
 pub mod export;
+pub mod imaging;
 pub mod maintenance;
+pub mod media_total;
 pub mod rate_limiter;
 pub mod sse_tickets;
+pub mod upload_admission;
+pub mod video;

backend/src/services/rate_limiter.rs

@@ -17,13 +17,14 @@ impl RateLimiter {
         }
     }
 
-    /// Returns `true` if the request is allowed, `false` if rate-limited.
-    pub fn check(&self, key: impl Into<String>, max: usize, window: Duration) -> bool {
-        self.check_with_retry(key, max, window).is_ok()
-    }
-
     /// Returns `Ok(())` if allowed, `Err(retry_after_secs)` if rate-limited.
     /// `retry_after_secs` is how long until the oldest slot in the window expires.
+    ///
+    /// This is deliberately the ONLY entry point. There used to be a `check()` wrapper
+    /// returning a plain bool, and 7 of the 8 call sites used it and then hard-coded
+    /// `None` for the response's `Retry-After` — so a throttled client was told to back
+    /// off but never for how long. Forcing every caller through the `Result` makes the
+    /// retry delay impossible to discard by accident.
     pub fn check_with_retry(
         &self,
         key: impl Into<String>,
@@ -39,8 +40,18 @@ impl RateLimiter {
             timestamps.push(now);
             Ok(())
         } else {
-            // The oldest timestamp expires at oldest + window; compute remaining seconds
-            let oldest = timestamps[0];
+            // The oldest timestamp expires at oldest + window; compute remaining seconds.
+            //
+            // `first()`, not `[0]`: with `max == 0` the length check above is false even on an
+            // empty vec, so indexing would panic — WHILE HOLDING THIS MUTEX. That poisons it
+            // process-wide, so every subsequent `.lock().unwrap()` panics too: upload, feed,
+            // join, recover, social, export and the hourly maintenance task all die, and only
+            // a container restart brings them back. `max == 0` is not reachable through the
+            // admin API (every numeric spec has min = 1) but a direct DB edit would do it, and
+            // the blast radius does not justify the sharper syntax.
+            let Some(&oldest) = timestamps.first() else {
+                return Ok(());
+            };
             let elapsed = now.duration_since(oldest);
             let remaining = window.saturating_sub(elapsed);
             Err(remaining.as_secs().max(1))
@@ -84,6 +95,11 @@ impl RateLimiter {
 /// appends is the real client. A client can prepend arbitrary spoofed values to
 /// the left of XFF to dodge throttles — those are ignored here. This assumes
 /// exactly one trusted proxy (Caddy); revisit if that changes.
+///
+/// Pass the peer address as `fallback`, never a constant. Every caller used to pass
+/// the literal `"unknown"`, so any request that arrived without XFF — i.e. anything
+/// reaching the app directly rather than through Caddy — shared ONE bucket with every
+/// other such request, turning the limiter into a self-inflicted global throttle.
 pub fn client_ip(headers: &axum::http::HeaderMap, fallback: &str) -> String {
     headers
         .get("x-forwarded-for")
@@ -104,29 +120,35 @@ mod tests {
     #[test]
     fn allows_up_to_max_then_blocks() {
         let rl = RateLimiter::new();
-        assert!(rl.check("k", 3, MIN));
-        assert!(rl.check("k", 3, MIN));
-        assert!(rl.check("k", 3, MIN));
-        assert!(!rl.check("k", 3, MIN), "the 4th request must be blocked");
+        assert!(rl.check_with_retry("k", 3, MIN).is_ok());
+        assert!(rl.check_with_retry("k", 3, MIN).is_ok());
+        assert!(rl.check_with_retry("k", 3, MIN).is_ok());
+        assert!(
+            rl.check_with_retry("k", 3, MIN).is_err(),
+            "the 4th request must be blocked"
+        );
     }
 
     #[test]
     fn keys_are_independent() {
         let rl = RateLimiter::new();
-        assert!(rl.check("a", 1, MIN));
-        assert!(!rl.check("a", 1, MIN));
-        assert!(rl.check("b", 1, MIN), "a different key has its own window");
+        assert!(rl.check_with_retry("a", 1, MIN).is_ok());
+        assert!(rl.check_with_retry("a", 1, MIN).is_err());
+        assert!(
+            rl.check_with_retry("b", 1, MIN).is_ok(),
+            "a different key has its own window"
+        );
     }
 
     #[test]
     fn window_slides_and_allows_again_after_expiry() {
         let rl = RateLimiter::new();
         let w = Duration::from_millis(40);
-        assert!(rl.check("k", 1, w));
-        assert!(!rl.check("k", 1, w));
+        assert!(rl.check_with_retry("k", 1, w).is_ok());
+        assert!(rl.check_with_retry("k", 1, w).is_err());
         std::thread::sleep(Duration::from_millis(55));
         assert!(
-            rl.check("k", 1, w),
+            rl.check_with_retry("k", 1, w).is_ok(),
             "the slot should expire once the window passes"
         );
     }
@@ -191,10 +213,13 @@ mod tests {
     #[test]
     fn clear_resets_every_window() {
         let rl = RateLimiter::new();
-        assert!(rl.check("k", 1, MIN));
-        assert!(!rl.check("k", 1, MIN));
+        assert!(rl.check_with_retry("k", 1, MIN).is_ok());
+        assert!(rl.check_with_retry("k", 1, MIN).is_err());
         rl.clear();
-        assert!(rl.check("k", 1, MIN), "clear() must free the window");
+        assert!(
+            rl.check_with_retry("k", 1, MIN).is_ok(),
+            "clear() must free the window"
+        );
     }
 
     /// `prune()` is a memory-leak guard: without it a long-lived process keeps one HashMap
@@ -216,7 +241,7 @@ mod tests {
             .insert("stale".to_string(), vec![ancient]);
 
         // ...alongside a key that is still inside its window.
-        assert!(rl.check("live", 5, MIN));
+        assert!(rl.check_with_retry("live", 5, MIN).is_ok());
         assert_eq!(rl.windows.lock().unwrap().len(), 2);
 
         rl.prune();
@@ -239,13 +264,13 @@ mod tests {
         // prune() dropped live keys, every background sweep would hand attackers a fresh
         // budget.
         let rl = RateLimiter::new();
-        assert!(rl.check("k", 1, MIN));
-        assert!(!rl.check("k", 1, MIN));
+        assert!(rl.check_with_retry("k", 1, MIN).is_ok());
+        assert!(rl.check_with_retry("k", 1, MIN).is_err());
 
         rl.prune();
 
         assert!(
-            !rl.check("k", 1, MIN),
+            rl.check_with_retry("k", 1, MIN).is_err(),
             "prune() must not clear a window that is still active"
         );
     }

backend/src/services/sse_tickets.rs

@@ -13,6 +13,36 @@ use rand::Rng;
 /// stream open. Tickets are consumed on use and expire after `TTL`.
 const TTL: Duration = Duration::from_secs(30);
 
+/// Ceiling on outstanding tickets across the whole process.
+///
+/// Not really about the bytes (~120 each) — about `issue` having had no bound of any kind.
+/// Sized well above a real event: ~1000 concurrent clients each holding one live 30 s ticket.
+const MAX_TICKETS: usize = 4096;
+
+/// Live tickets one session may hold. Above 1 because two tabs sharing a token open their
+/// EventSources concurrently; 4 absorbs that without letting a reconnect loop accumulate.
+const MAX_TICKETS_PER_SESSION: usize = 4;
+
+/// What a ticket may be redeemed for.
+///
+/// The store began life serving only SSE and stayed untyped when the export download started
+/// reusing it, which silently made the two interchangeable. That is not a theoretical mixing
+/// concern: `POST /stream/ticket` is rate-limited at 60/min per user and charges nothing, while
+/// `POST /export/ticket` charges one of three PER-DAY downloads. An untyped ticket let any guest
+/// mint at the cheap endpoint and redeem at the expensive one, so the daily export limit was
+/// bypassable ~60×/minute — each redemption streaming the whole multi-GB keepsake, `no-store`,
+/// off the same filesystem Postgres writes WAL to.
+///
+/// `consume` therefore requires the kind to MATCH. A ticket is only ever valid for the thing it
+/// was minted for.
+#[derive(Clone, Copy, PartialEq, Eq, Debug)]
+pub enum TicketKind {
+    /// Opens the SSE stream (`GET /stream`). Cheap, high volume.
+    Sse,
+    /// Downloads an export archive (`GET /export/{zip,html}`). Expensive, rate-limited per day.
+    Download,
+}
+
 #[derive(Clone)]
 pub struct SseTicketStore {
     inner: Arc<Mutex<HashMap<String, Entry>>>,
@@ -22,6 +52,7 @@ pub struct SseTicketStore {
 struct Entry {
     token_hash: String,
     issued_at: Instant,
+    kind: TicketKind,
 }
 
 impl SseTicketStore {
@@ -39,28 +70,80 @@ impl SseTicketStore {
     }
 
     /// Mint a new ticket bound to the caller's session (identified by token hash).
-    pub fn issue(&self, token_hash: String) -> String {
+    ///
+    /// `None` when the store is at capacity — the caller should answer 503, not evict.
+    ///
+    /// Three bounds, because `issue` had none: no size cap, no per-caller cap, and no rate
+    /// limit on the endpoint, while `prune` ran only hourly against a 30-second TTL. So any
+    /// authenticated session could loop the endpoint and grow the map for an hour.
+    pub fn issue(&self, token_hash: String, kind: TicketKind) -> Option<String> {
         let ticket = random_ticket();
         let mut map = self.inner.lock().unwrap();
+
+        // Prune on issue rather than only hourly. This alone changes the bound from "tickets
+        // minted since the last maintenance tick" to "tickets live at once", which is what the
+        // 30 s TTL was always meant to express.
+        map.retain(|_, e| e.issued_at.elapsed() <= TTL);
+
+        // Cap the caller's own outstanding tickets, evicting their oldest. NOT one-per-session:
+        // two tabs sharing a token open their EventSources concurrently, and having tab B
+        // invalidate tab A's unconsumed ticket looks exactly like a flaky SSE connection.
+        let mut mine: Vec<(String, Instant)> = map
+            .iter()
+            .filter(|(_, e)| e.token_hash == token_hash)
+            .map(|(k, e)| (k.clone(), e.issued_at))
+            .collect();
+        if mine.len() >= MAX_TICKETS_PER_SESSION {
+            mine.sort_by_key(|(_, issued)| *issued);
+            for (key, _) in mine.iter().take(mine.len() - MAX_TICKETS_PER_SESSION + 1) {
+                map.remove(key);
+            }
+        }
+
+        // At capacity, REFUSE — never evict a stranger's ticket. Evicting would let one
+        // misbehaving client deny SSE to the whole venue, which is worse than failing the
+        // request that hit the ceiling.
+        if map.len() >= MAX_TICKETS {
+            tracing::warn!(
+                outstanding = map.len(),
+                "SSE ticket store at capacity; refusing to mint"
+            );
+            return None;
+        }
+
         map.insert(
             ticket.clone(),
             Entry {
                 token_hash,
                 issued_at: Instant::now(),
+                kind,
             },
         );
-        ticket
+        Some(ticket)
     }
 
-    /// Consume a ticket. Returns `Some(token_hash)` if the ticket exists and is
-    /// not expired. Single-use: the ticket is removed regardless of whether it
-    /// was still fresh, so a replay can't slip through after expiry.
-    pub fn consume(&self, ticket: &str) -> Option<String> {
+    /// Consume a ticket minted for `kind`. Returns `Some(token_hash)` if the ticket exists, is
+    /// not expired, and was minted for this purpose. Single-use: the ticket is removed regardless
+    /// of whether it was still fresh, so a replay can't slip through after expiry.
+    ///
+    /// A ticket of the WRONG kind is also removed. It was a valid ticket the caller legitimately
+    /// held, so this is not punitive — but leaving it would let a redemption loop probe the store
+    /// without ever spending anything, and the client has no legitimate reason to present a
+    /// ticket at the wrong endpoint.
+    pub fn consume(&self, ticket: &str, kind: TicketKind) -> Option<String> {
         let mut map = self.inner.lock().unwrap();
         let entry = map.remove(ticket)?;
         if entry.issued_at.elapsed() > TTL {
             return None;
         }
+        if entry.kind != kind {
+            tracing::warn!(
+                expected = ?kind,
+                found = ?entry.kind,
+                "ticket presented at the wrong endpoint; rejected"
+            );
+            return None;
+        }
         Some(entry.token_hash)
     }
 
@@ -84,14 +167,57 @@ fn random_ticket() -> String {
 mod tests {
     use super::*;
 
+    /// `issue` now returns `Option`; in every test below the store is far from capacity, so an
+    /// `expect` here documents that refusing is exceptional rather than routine.
+    fn issue(store: &SseTicketStore, hash: &str) -> String {
+        store
+            .issue(hash.into(), TicketKind::Sse)
+            .expect("store has capacity")
+    }
+
+    /// The store is shared by two endpoints with wildly different costs: `/stream/ticket` is
+    /// 60/min per user and free, `/export/ticket` charges one of three PER-DAY downloads. While
+    /// entries were untyped, a ticket minted at the cheap endpoint opened the expensive one — so
+    /// the daily export limit could be bypassed ~60×/minute, each redemption streaming the whole
+    /// multi-GB keepsake off the disk Postgres writes WAL to.
+    ///
+    /// Asserted in BOTH directions so this cannot be "fixed" by a check that only guards one.
+    #[test]
+    fn a_ticket_is_only_valid_for_the_purpose_it_was_minted_for() {
+        let store = SseTicketStore::new();
+
+        let sse = store.issue("h".into(), TicketKind::Sse).unwrap();
+        assert_eq!(
+            store.consume(&sse, TicketKind::Download),
+            None,
+            "an SSE ticket must not open the export download"
+        );
+
+        let dl = store.issue("h".into(), TicketKind::Download).unwrap();
+        assert_eq!(
+            store.consume(&dl, TicketKind::Sse),
+            None,
+            "a download ticket must not open the SSE stream"
+        );
+
+        // And the matching cases still work, so the guard is not simply rejecting everything.
+        let sse = store.issue("h".into(), TicketKind::Sse).unwrap();
+        assert_eq!(store.consume(&sse, TicketKind::Sse).as_deref(), Some("h"));
+        let dl = store.issue("h".into(), TicketKind::Download).unwrap();
+        assert_eq!(
+            store.consume(&dl, TicketKind::Download).as_deref(),
+            Some("h")
+        );
+    }
+
     #[test]
     fn issue_then_consume_returns_the_hash_exactly_once() {
         let store = SseTicketStore::new();
-        let ticket = store.issue("hash-1".into());
-        assert_eq!(store.consume(&ticket).as_deref(), Some("hash-1"));
+        let ticket = issue(&store, "hash-1");
+        assert_eq!(store.consume(&ticket, TicketKind::Sse).as_deref(), Some("hash-1"));
         // Single-use: a replay of the same ticket is rejected.
         assert_eq!(
-            store.consume(&ticket),
+            store.consume(&ticket, TicketKind::Sse),
             None,
             "a consumed ticket must not be reusable"
         );
@@ -100,14 +226,14 @@ mod tests {
     #[test]
     fn unknown_ticket_consumes_to_none() {
         let store = SseTicketStore::new();
-        assert_eq!(store.consume("never-issued"), None);
+        assert_eq!(store.consume("never-issued", TicketKind::Sse), None);
     }
 
     #[test]
     fn issued_tickets_are_unique_and_hex() {
         let store = SseTicketStore::new();
-        let a = store.issue("h".into());
-        let b = store.issue("h".into());
+        let a = issue(&store, "h");
+        let b = issue(&store, "h");
         assert_ne!(a, b, "each ticket must be unique");
         assert_eq!(a.len(), 48, "24 random bytes → 48 hex chars");
         assert!(a.chars().all(|c| c.is_ascii_hexdigit()));
@@ -116,29 +242,106 @@ mod tests {
     #[test]
     fn fresh_ticket_survives_prune() {
         let store = SseTicketStore::new();
-        let ticket = store.issue("h".into());
+        let ticket = issue(&store, "h");
         store.prune(); // not expired → kept
-        assert_eq!(store.consume(&ticket).as_deref(), Some("h"));
+        assert_eq!(store.consume(&ticket, TicketKind::Sse).as_deref(), Some("h"));
     }
 
-    #[test]
-    fn expired_ticket_consumes_to_none() {
-        // Construct an entry that is already past the TTL and confirm consume() rejects it.
-        let store = SseTicketStore::new();
-        let stale = "stale-ticket".to_string();
+    /// Build an entry that is already past the TTL.
+    fn insert_stale(store: &SseTicketStore, key: &str, token_hash: &str) {
         store.inner.lock().unwrap().insert(
-            stale.clone(),
+            key.to_string(),
             Entry {
-                token_hash: "h".into(),
+                kind: TicketKind::Sse,
+                token_hash: token_hash.into(),
                 issued_at: Instant::now()
                     .checked_sub(TTL + Duration::from_secs(1))
                     .expect("host uptime should exceed the ticket TTL"),
             },
         );
+    }
+
+    #[test]
+    fn expired_ticket_consumes_to_none() {
+        let store = SseTicketStore::new();
+        insert_stale(&store, "stale-ticket", "h");
         assert_eq!(
-            store.consume(&stale),
+            store.consume("stale-ticket", TicketKind::Sse),
             None,
             "an expired ticket must not authenticate"
         );
     }
+
+    /// The TTL is 30 s but `prune` only ran hourly, so the map was really bounded by "tickets
+    /// minted in the last hour" — which is unbounded for a client in a loop.
+    #[test]
+    fn issuing_prunes_expired_entries() {
+        let store = SseTicketStore::new();
+        insert_stale(&store, "stale-a", "someone-else");
+        insert_stale(&store, "stale-b", "someone-else");
+        issue(&store, "h");
+        assert_eq!(
+            store.inner.lock().unwrap().len(),
+            1,
+            "issue must reclaim expired slots, not merely add to them"
+        );
+    }
+
+    /// Two tabs sharing a token is normal, so the per-session cap must be above 1 — but a
+    /// reconnect loop must not accumulate. The caller's OWN oldest is what gets evicted.
+    #[test]
+    fn a_session_is_capped_and_evicts_only_its_own_oldest() {
+        let store = SseTicketStore::new();
+        let stranger = issue(&store, "other-session");
+
+        let mut mine: Vec<String> = Vec::new();
+        for _ in 0..MAX_TICKETS_PER_SESSION + 2 {
+            mine.push(issue(&store, "mine"));
+        }
+
+        let live = mine
+            .iter()
+            .filter(|t| store.inner.lock().unwrap().contains_key(*t))
+            .count();
+        assert_eq!(live, MAX_TICKETS_PER_SESSION, "one session, bounded");
+        assert!(
+            store.inner.lock().unwrap().contains_key(&mine[mine.len() - 1]),
+            "the newest ticket is the one the caller is about to use"
+        );
+        assert_eq!(
+            store.consume(&stranger, TicketKind::Sse).as_deref(),
+            Some("other-session"),
+            "another session's ticket must survive — evicting it would let one client deny \
+             SSE to the venue"
+        );
+    }
+
+    /// At capacity the store REFUSES rather than evicting a stranger. Refusing fails the one
+    /// request that hit the ceiling; evicting would break an unrelated client's live stream.
+    #[test]
+    fn at_capacity_the_store_refuses_instead_of_evicting() {
+        let store = SseTicketStore::new();
+        {
+            let mut map = store.inner.lock().unwrap();
+            for i in 0..MAX_TICKETS {
+                map.insert(
+                    format!("filler-{i}"),
+                    Entry {
+                        kind: TicketKind::Sse,
+                        token_hash: format!("session-{i}"),
+                        issued_at: Instant::now(),
+                    },
+                );
+            }
+        }
+        assert_eq!(
+            store.issue("newcomer".into(), TicketKind::Sse),
+            None,
+            "a full store must refuse, so the caller can answer 503"
+        );
+        assert!(
+            store.inner.lock().unwrap().contains_key("filler-0"),
+            "no existing ticket may be sacrificed to make room"
+        );
+    }
 }

backend/src/services/upload_admission.rs

@@ -0,0 +1,155 @@
+//! Admission control for upload bodies, budgeted in BYTES rather than requests.
+//!
+//! ## Why this has to exist
+//!
+//! The keepsake headroom gate in `handlers::upload` cannot bound a burst, and the reason is
+//! structural rather than a bug in the gate: the request body is streamed to a temp file during
+//! multipart parsing, so the bytes are already on disk by the time any check runs. The gate can
+//! only refuse to COMMIT them. Nothing upstream limited how many bodies stream at once — axum has
+//! no such limit, the tower stack is just `TraceLayer`, and Caddy passes requests straight
+//! through.
+//!
+//! So the failure mode is the ordinary one, not an attack: the ceremony ends, ~100 guests tap
+//! "upload all", and ~100 bodies stream concurrently. At phone-video sizes that is 10-20 GB of
+//! `.tmp` files on a 40 GB volume, none of it visible to the gate, and `DISK_RESERVE_BYTES` — the
+//! 10 GB standing between the party and Postgres losing the volume it writes WAL to — is consumed
+//! by transient files. The `.tmp` sweeper only reclaims files idle for an hour, correctly, which
+//! means nothing reclaims a burst on this timescale.
+//!
+//! ## Why bytes and not a request count
+//!
+//! A flat "N concurrent uploads" limit has to be sized for the worst case (a 500 MB video), which
+//! makes it absurdly restrictive for the common case (a 3 MB photo). Budgeting bytes lets one
+//! 500 MB video and two hundred photos coexist under the same ceiling, and it means the ceiling is
+//! stated in the unit the disk actually cares about.
+//!
+//! The reservation is the streaming CAP, not the real size — the real size is unknowable until the
+//! body has been read, which is far too late. Reserving the cap is deliberately pessimistic; that
+//! pessimism is the safety margin.
+//!
+//! ## Why a permit and not a counter
+//!
+//! `OwnedSemaphorePermit` releases on drop. Every path out of the upload handler — success, error,
+//! a client vanishing mid-body, a panic — therefore returns the reservation without any explicit
+//! bookkeeping. A hand-rolled `AtomicI64` would need a decrement on each of those paths, and the
+//! one that gets missed is the one that leaks the budget until restart.
+
+use std::sync::Arc;
+use std::time::Duration;
+
+use tokio::sync::{OwnedSemaphorePermit, Semaphore};
+
+/// Total transient upload bytes allowed on disk at once, in MiB.
+///
+/// Sized against `DISK_RESERVE_BYTES` (10 GB): the reserve must survive a full burst with room to
+/// spare, since Postgres is writing WAL to the same filesystem throughout. 4 GiB leaves ~6 GB of
+/// the reserve untouched at the worst moment.
+///
+/// It is NOT a throughput limit. On 2 vCPU the box cannot usefully ingest more than this at once
+/// anyway — compression, ffmpeg, Postgres and TLS all contend for the same two cores — so the
+/// budget mostly converts "everything is slow and the disk fills" into "a few uploads wait".
+const BUDGET_MIB: u32 = 4096;
+
+/// How long an upload waits for room before being told to come back.
+///
+/// Long enough to absorb the burst (a photo holds its reservation for well under a second), short
+/// enough that a guest is not left staring at a spinner. On timeout the handler answers 503 with
+/// `Retry-After`, which the client queue already treats as transient and retries with backoff.
+const WAIT: Duration = Duration::from_secs(20);
+
+#[derive(Clone)]
+pub struct UploadAdmission {
+    permits: Arc<Semaphore>,
+}
+
+impl UploadAdmission {
+    pub fn new() -> Self {
+        Self {
+            permits: Arc::new(Semaphore::new(BUDGET_MIB as usize)),
+        }
+    }
+
+    /// Reserve room for a body capped at `cap_bytes`. The returned permit must be held for as long
+    /// as the temp file exists.
+    ///
+    /// `None` means the wait timed out and the caller should shed the request.
+    ///
+    /// A cap larger than the whole budget is clamped rather than refused. Otherwise an operator
+    /// raising `max_video_size_mb` above the budget would make `acquire_many` unsatisfiable and
+    /// every video upload would hang until timeout — a config change silently disabling video for
+    /// the event. Clamped, such an upload simply gets the whole budget to itself, which is the
+    /// honest interpretation of "one file may fill the machine".
+    pub async fn reserve(&self, cap_bytes: usize) -> Option<OwnedSemaphorePermit> {
+        let mib = cap_bytes.div_ceil(1024 * 1024).max(1);
+        let want = u32::try_from(mib).unwrap_or(BUDGET_MIB).min(BUDGET_MIB);
+        match tokio::time::timeout(
+            WAIT,
+            self.permits.clone().acquire_many_owned(want),
+        )
+        .await
+        {
+            Ok(Ok(permit)) => Some(permit),
+            // The semaphore is never closed, so `Err` here is unreachable in practice; treat it
+            // the same as a timeout rather than panicking on the upload path.
+            Ok(Err(_)) => None,
+            Err(_) => {
+                tracing::warn!(
+                    requested_mib = want,
+                    "upload admission timed out; shedding to keep transient temp files bounded"
+                );
+                None
+            }
+        }
+    }
+}
+
+impl Default for UploadAdmission {
+    fn default() -> Self {
+        Self::new()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// The budget must actually block once exhausted — otherwise this whole module is decoration.
+    #[tokio::test]
+    async fn a_full_budget_sheds_instead_of_admitting() {
+        let admission = UploadAdmission::new();
+        let whole = admission
+            .reserve(BUDGET_MIB as usize * 1024 * 1024)
+            .await
+            .expect("first reservation takes the whole budget");
+
+        // Nothing left: a second reservation must not be granted. Raced against a short timeout so
+        // the test does not sit for the full WAIT.
+        let blocked = tokio::time::timeout(
+            Duration::from_millis(150),
+            admission.reserve(1024 * 1024),
+        )
+        .await;
+        assert!(blocked.is_err(), "budget exhausted, yet a reservation was granted");
+
+        // ...and releasing the permit makes room again, so the budget is not a one-way latch.
+        drop(whole);
+        assert!(
+            admission.reserve(1024 * 1024).await.is_some(),
+            "budget did not recover after the permit was dropped"
+        );
+    }
+
+    /// A cap above the whole budget must be clamped, not left unsatisfiable. Unclamped,
+    /// `acquire_many` for more permits than exist never completes, so raising
+    /// `max_video_size_mb` past the budget would silently hang every video upload for 20s and
+    /// then shed it.
+    #[tokio::test]
+    async fn a_cap_larger_than_the_budget_is_clamped_rather_than_unsatisfiable() {
+        let admission = UploadAdmission::new();
+        let oversized = (BUDGET_MIB as usize + 4096) * 1024 * 1024;
+        assert!(
+            admission.reserve(oversized).await.is_some(),
+            "an over-budget cap must still be admittable on an idle server"
+        );
+    }
+}

backend/src/services/video.rs

@@ -0,0 +1,263 @@
+//! Poster-frame extraction, shared by the compression worker and the HTML export.
+//!
+//! Both used to spawn `ffmpeg` themselves with the same broken invocation:
+//!
+//! ```text
+//! ffmpeg -i <src> -vframes 1 -ss 00:00:01 -vf scale=… -y <out>
+//! ```
+//!
+//! `-ss` AFTER `-i` is an output-side seek. Against a clip of a second or less ffmpeg exits **0 and
+//! writes nothing** — and both call sites gated on the exit status, so neither noticed. The worker
+//! then wrote `thumbnail_path` for a file that was never created (404 in the live feed) and the
+//! export listed the entry in `data.json` while the ZIP writer skipped it (a broken image tile in
+//! the keepsake). Every server-side signal stayed green. Phones produce such clips constantly:
+//! mis-taps, Live Photos, boomerangs.
+//!
+//! This module exists for the same reason `imaging.rs` does — that one was created when compression
+//! and export duplicated decode logic, and it paid off immediately when the `max_alloc` fix landed
+//! in both workers at once. Same duplication, same fix.
+
+use std::path::Path;
+use std::time::Duration;
+
+use anyhow::{Context, Result};
+
+/// A malformed video can hang `ffmpeg` indefinitely. In the compression worker that never releases
+/// the semaphore permit and the pool eventually deadlocks; in the export worker it strands the job
+/// at `running` so the keepsake never completes. `export.rs` had NO timeout at all before this
+/// module — sharing the spawn fixes that too.
+/// 45s, not the 120s this started at. The timeout is not a budget for honest work — a poster
+/// frame from a phone clip takes well under a second, and `-ss` before `-i` means even a 500 MB
+/// file seeks rather than scans. It is purely the ceiling on how long a pathological input may
+/// hold a compression permit that guests' photos are queued behind, so it should be as tight as
+/// it can be without ever cutting off real work.
+const FFMPEG_TIMEOUT: Duration = Duration::from_secs(45);
+
+/// Seek positions to try, in order.
+///
+/// One second first: the opening frame of a real video is often black, a fade-in, or motion-blurred
+/// as the camera settles, so it makes a poor poster. Zero second as the fallback, which is what
+/// makes short clips work — and it is genuinely required, not defensive. Moving `-ss` before `-i`
+/// (an input-side seek) is necessary but NOT sufficient: seeking to 1 s in a 1.000 s clip is still
+/// past the last frame, and ffmpeg still exits 0 having written nothing. Verified against the real
+/// production image.
+const SEEK_POSITIONS: &[&str] = &["00:00:01", "0"];
+
+/// Extract one poster frame from `src` into `dest`, scaled to `width` px wide.
+///
+/// `Ok(false)` means the video yielded no frame — a normal outcome for a very short or unusual
+/// clip, NOT an error. Callers must degrade (no poster) rather than fail the upload: treating this
+/// as an error would soft-delete every sub-second video, turning a cosmetic defect into data loss.
+///
+/// `Err` is reserved for something genuinely wrong — a hang we had to kill, or a failure to spawn.
+pub async fn extract_poster_frame(src: &Path, dest: &Path, width: u32) -> Result<bool> {
+    for seek in SEEK_POSITIONS {
+        // A stale file from a previous attempt would be indistinguishable from a fresh success.
+        let _ = tokio::fs::remove_file(dest).await;
+
+        run_ffmpeg(src, dest, width, seek).await?;
+
+        // THE CHECK BOTH CALL SITES WERE MISSING: ask the filesystem, not the exit status.
+        // Non-empty, because a zero-byte file is not a poster either.
+        if tokio::fs::metadata(dest)
+            .await
+            .map(|m| m.is_file() && m.len() > 0)
+            .unwrap_or(false)
+        {
+            return Ok(true);
+        }
+    }
+
+    // Leave nothing behind for a caller to mistake for a result.
+    let _ = tokio::fs::remove_file(dest).await;
+    Ok(false)
+}
+
+/// Run one ffmpeg attempt. A non-zero exit is NOT an error here — the artifact check above is the
+/// authority, and a corrupt input that fails at 1 s may still yield a frame at 0.
+async fn run_ffmpeg(src: &Path, dest: &Path, width: u32, seek: &str) -> Result<()> {
+    let child = tokio::process::Command::new("ffmpeg")
+        .args([
+            // BEFORE -i: an input-side seek. See SEEK_POSITIONS.
+            "-ss",
+            seek,
+            "-i",
+            src.to_str().unwrap_or_default(),
+            "-vframes",
+            "1",
+            "-vf",
+            &format!("scale={width}:-1"),
+            "-y",
+            dest.to_str().unwrap_or_default(),
+        ])
+        // ffmpeg writes the poster to `dest` itself; nothing here ever reads stdout, so
+        // giving it a pipe only created something that could fill.
+        .stdout(std::process::Stdio::null())
+        // stderr IS piped — it is the only diagnostic when a clip yields no frame — but it
+        // must be DRAINED, which is the whole point of `wait_with_output` below.
+        .stderr(std::process::Stdio::piped())
+        .kill_on_drop(true)
+        .spawn()
+        .context("failed to spawn ffmpeg")?;
+
+    // `wait_with_output`, NOT `wait`. ffmpeg is verbose on stderr (banner, stream info,
+    // per-frame progress) and `wait()` reads neither pipe — so once the ~64 KiB pipe buffer
+    // filled, ffmpeg blocked writing, `wait()` never returned, and the call burned the full
+    // timeout. That is not merely slow: the timeout is an `Err`, so after 2 seek positions x
+    // 3 compression attempts the caller soft-deletes a perfectly playable video for a
+    // poster-frame failure. `wait_with_output` polls the pipe and the exit status together.
+    //
+    // It also CONSUMES the child, so the explicit `child.kill()` that used to sit on the
+    // timeout arm cannot exist here — and is not needed: `kill_on_drop(true)` is set above,
+    // and dropping the future on timeout drops the child with it.
+    let out = match tokio::time::timeout(FFMPEG_TIMEOUT, child.wait_with_output()).await {
+        Ok(res) => res.context("ffmpeg wait failed")?,
+        Err(_) => anyhow::bail!("ffmpeg timed out after {}s", FFMPEG_TIMEOUT.as_secs()),
+    };
+
+    // A non-zero exit is not an error (see the doc comment) — the artifact check in
+    // `extract_poster_frame` is the authority. Log the tail so a systematically failing
+    // format is diagnosable without turning it into data loss.
+    if !out.status.success() {
+        tracing::debug!(
+            seek,
+            status = ?out.status,
+            stderr = %tail_lines(&out.stderr, 10),
+            "ffmpeg exited non-zero; the artifact check decides"
+        );
+    }
+    Ok(())
+}
+
+/// Last `n` lines of a child's stderr, lossily decoded.
+///
+/// Bounded on purpose: ffmpeg's stderr is unbounded, and the reason we now drain it is that
+/// unbounded output used to be a hazard. Emitting all of it into a log line — into container
+/// logs that are themselves size-capped — would just move the problem.
+fn tail_lines(bytes: &[u8], n: usize) -> String {
+    let text = String::from_utf8_lossy(bytes);
+    let lines: Vec<&str> = text.lines().filter(|l| !l.trim().is_empty()).collect();
+    lines[lines.len().saturating_sub(n)..].join(" | ")
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Is there a usable `ffmpeg` on PATH?
+    ///
+    /// The poster-frame path shells out, and `extract_poster_frame` documents `Err` as meaning
+    /// "a hang or a SPAWN failure" — which is exactly what a missing binary produces. So on a
+    /// machine without ffmpeg the test below stops exercising the case it names (missing INPUT)
+    /// and instead reports a code defect that isn't there. The runtime image installs ffmpeg
+    /// (see backend/Dockerfile), so this only ever skips on a bare developer machine.
+    fn ffmpeg_available() -> bool {
+        std::process::Command::new("ffmpeg")
+            .arg("-version")
+            .stdout(std::process::Stdio::null())
+            .stderr(std::process::Stdio::null())
+            .status()
+            .is_ok()
+    }
+
+    /// The order is the whole fix. `-ss` must precede `-i`, and 0 must be tried after 1 s.
+    #[test]
+    fn the_fallback_seek_exists_and_comes_last() {
+        assert_eq!(
+            SEEK_POSITIONS,
+            &["00:00:01", "0"],
+            "1s first for a better poster, 0 as the fallback that makes short clips work"
+        );
+    }
+
+    /// A missing input yields no frame rather than an error: the caller must degrade to "no
+    /// poster", never fail the upload. `Err` is reserved for a hang or a spawn failure.
+    #[tokio::test]
+    async fn a_missing_source_yields_no_frame_rather_than_an_error() {
+        if !ffmpeg_available() {
+            eprintln!(
+                "SKIP a_missing_source_yields_no_frame_rather_than_an_error: no ffmpeg on PATH. \
+                 A missing binary is a spawn failure, which this function returns Err for by \
+                 design, so the missing-INPUT case cannot be exercised here. Install ffmpeg to \
+                 run it (the runtime image already has it)."
+            );
+            return;
+        }
+        let dir = std::env::temp_dir().join(format!("es-video-{}", std::process::id()));
+        std::fs::create_dir_all(&dir).unwrap();
+        let dest = dir.join("out.jpg");
+
+        let got = extract_poster_frame(Path::new("/nonexistent/clip.mp4"), &dest, 400).await;
+
+        match got {
+            Ok(false) => {}
+            other => panic!("expected Ok(false) for a missing input, got {other:?}"),
+        }
+        assert!(
+            !dest.exists(),
+            "a failed extraction must leave nothing a caller could mistake for a poster"
+        );
+        let _ = std::fs::remove_dir_all(&dir);
+    }
+
+    #[test]
+    fn the_stderr_tail_is_bounded_and_survives_invalid_utf8() {
+        let noisy: Vec<u8> = (0..500)
+            .map(|i| format!("line {i}\n"))
+            .collect::<String>()
+            .into_bytes();
+        let got = tail_lines(&noisy, 3);
+        assert_eq!(got, "line 497 | line 498 | line 499");
+
+        // ffmpeg emits filenames verbatim, so its stderr is not guaranteed to be UTF-8.
+        assert_eq!(tail_lines(&[b'o', b'k', 0xff], 5), "ok\u{fffd}");
+        assert_eq!(tail_lines(b"", 5), "");
+    }
+
+    /// A real extraction must finish in a small fraction of `FFMPEG_TIMEOUT`.
+    ///
+    /// Wall-clock is the ONLY observable of the bug this guards: piping stderr and then
+    /// calling `wait()` (which drains nothing) blocks ffmpeg on a full pipe buffer until the
+    /// timeout fires, and the timeout is an `Err`, so the upload is soft-deleted. The
+    /// assertion is deliberately on elapsed time, not on the exit status.
+    ///
+    /// Honest limitation: our fixture is quiet enough not to fill a 64 KiB pipe on its own,
+    /// so this catches a regression to `wait()` only in combination with a verbose input. It
+    /// is still worth pinning — a reverted drain plus any chatty clip is data loss.
+    #[tokio::test]
+    async fn a_real_clip_yields_a_poster_well_inside_the_timeout() {
+        if tokio::process::Command::new("ffmpeg")
+            .arg("-version")
+            .stdout(std::process::Stdio::null())
+            .stderr(std::process::Stdio::null())
+            .status()
+            .await
+            .is_err()
+        {
+            eprintln!("skipping: ffmpeg not on PATH");
+            return;
+        }
+        let src = Path::new("../e2e/fixtures/media/sample.mp4");
+        if !src.exists() {
+            eprintln!("skipping: {} missing", src.display());
+            return;
+        }
+
+        let dir = std::env::temp_dir().join(format!("es-video-ok-{}", std::process::id()));
+        std::fs::create_dir_all(&dir).unwrap();
+        let dest = dir.join("poster.jpg");
+
+        let started = std::time::Instant::now();
+        let got = extract_poster_frame(src, &dest, 400).await;
+        let elapsed = started.elapsed();
+
+        assert!(matches!(got, Ok(true)), "expected a poster, got {got:?}");
+        assert!(dest.metadata().unwrap().len() > 0);
+        assert!(
+            elapsed < FFMPEG_TIMEOUT / 4,
+            "extraction took {elapsed:?}; a drained stderr finishes in well under \
+             {FFMPEG_TIMEOUT:?} — this is the pipe-deadlock regression guard"
+        );
+        let _ = std::fs::remove_dir_all(&dir);
+    }
+}

backend/src/state.rs

@@ -5,8 +5,10 @@ use crate::config::AppConfig;
 use crate::services::compression::CompressionWorker;
 use crate::services::config::ConfigCache;
 use crate::services::disk::DiskCache;
+use crate::services::media_total::MediaTotalCache;
 use crate::services::rate_limiter::RateLimiter;
 use crate::services::sse_tickets::SseTicketStore;
+use crate::services::upload_admission::UploadAdmission;
 
 #[derive(Clone, Debug)]
 pub struct SseEvent {
@@ -38,6 +40,12 @@ pub struct AppState {
     pub config_cache: ConfigCache,
     /// Cached total/free bytes for the media filesystem (quota + admin stats).
     pub disk_cache: DiskCache,
+    /// Cached sum of all media bytes, for the upload gate's keepsake-headroom check.
+    pub media_total: MediaTotalCache,
+    /// Byte budget for upload bodies currently streaming to temp files. The headroom gate can
+    /// only refuse to COMMIT bytes that are already on disk; this is what bounds how many get
+    /// there at once.
+    pub upload_admission: UploadAdmission,
 }
 
 impl AppState {
@@ -63,6 +71,8 @@ impl AppState {
             sse_tickets: SseTicketStore::new(),
             config_cache,
             disk_cache: DiskCache::new(),
+            media_total: MediaTotalCache::new(),
+            upload_admission: UploadAdmission::new(),
         }
     }
 }

backend/static/export-viewer/_app/env.js

@@ -1 +0,0 @@
-export const env={}

backend/static/export-viewer/_app/immutable/chunks/BJ__EZ0W.js

@@ -1 +0,0 @@
-import{u as o,n as t,o as c}from"./CcONa1Mr.js";function u(e){throw new Error("https://svelte.dev/e/lifecycle_outside_component")}function r(e){t===null&&u(),o(()=>{const n=c(e);if(typeof n=="function")return n})}export{r as o};

backend/static/export-viewer/_app/immutable/chunks/BRDva_z9.js

@@ -1 +0,0 @@
-import{f as l,g as o,p as u,i as n,j as d,k as m,h as p,e as _,m as v,l as k}from"./CcONa1Mr.js";class w{anchor;#t=new Map;#s=new Map;#e=new Map;#i=new Set;#f=!0;constructor(t,s=!0){this.anchor=t,this.#f=s}#a=t=>{if(this.#t.has(t)){var s=this.#t.get(t),e=this.#s.get(s);if(e)l(e),this.#i.delete(s);else{var f=this.#e.get(s);f&&(this.#s.set(s,f.effect),this.#e.delete(s),f.fragment.lastChild.remove(),this.anchor.before(f.fragment),e=f.effect)}for(const[i,a]of this.#t){if(this.#t.delete(i),i===t)break;const r=this.#e.get(a);r&&(o(r.effect),this.#e.delete(a))}for(const[i,a]of this.#s){if(i===s||this.#i.has(i))continue;const r=()=>{if(Array.from(this.#t.values()).includes(i)){var c=document.createDocumentFragment();v(a,c),c.append(n()),this.#e.set(i,{effect:a,fragment:c})}else o(a);this.#i.delete(i),this.#s.delete(i)};this.#f||!e?(this.#i.add(i),u(a,r,!1)):r()}}};#r=t=>{this.#t.delete(t);const s=Array.from(this.#t.values());for(const[e,f]of this.#e)s.includes(e)||(o(f.effect),this.#e.delete(e))};ensure(t,s){var e=m,f=k();if(s&&!this.#s.has(t)&&!this.#e.has(t))if(f){var i=document.createDocumentFragment(),a=n();i.append(a),this.#e.set(t,{effect:d(()=>s(a)),fragment:i})}else this.#s.set(t,d(()=>s(this.anchor)));if(this.#t.set(e,t),f){for(const[r,h]of this.#s)r===t?e.unskip_effect(h):e.skip_effect(h);for(const[r,h]of this.#e)r===t?e.unskip_effect(h.effect):e.skip_effect(h.effect);e.oncommit(this.#a),e.ondiscard(this.#r)}else p&&(this.anchor=_),this.#a(e)}}export{w as B};

backend/static/export-viewer/_app/immutable/chunks/Bxn6SmGf.js

@@ -1 +0,0 @@
-import{b as c,h as o,a as l,E as b,r as p,s as v,c as g,d,e as m}from"./CcONa1Mr.js";import{B as y}from"./BRDva_z9.js";function k(f,h,_=!1){var n;o&&(n=m,l());var s=new y(f),u=_?b:0;function t(a,r){if(o){var e=p(n);if(a!==parseInt(e.substring(1))){var i=v();g(i),s.anchor=i,d(!1),s.ensure(a,r),d(!0);return}}s.ensure(a,r)}c(()=>{var a=!1;h((r,e=0)=>{a=!0,t(e,r)}),a||t(-1,null)},u)}export{k as i};

backend/static/export-viewer/_app/immutable/chunks/RsTAN2PN.js

@@ -1 +0,0 @@
-import{A as v,i as d,B as l,C as u,D as T,T as p,F as h,h as i,e as s,R as E,a as y,G as g,c as w,H as N}from"./CcONa1Mr.js";const A=globalThis?.window?.trustedTypes&&globalThis.window.trustedTypes.createPolicy("svelte-trusted-html",{createHTML:t=>t});function M(t){return A?.createHTML(t)??t}function x(t){var r=v("template");return r.innerHTML=M(t.replaceAll("<!>","<!---->")),r.content}function n(t,r){var e=l;e.nodes===null&&(e.nodes={start:t,end:r,a:null,t:null})}function b(t,r){var e=(r&p)!==0,f=(r&h)!==0,a,_=!t.startsWith("<!>");return()=>{if(i)return n(s,null),s;a===void 0&&(a=x(_?t:"<!>"+t),e||(a=u(a)));var o=f||T?document.importNode(a,!0):a.cloneNode(!0);if(e){var c=u(o),m=o.lastChild;n(c,m)}else n(o,o);return o}}function C(t=""){if(!i){var r=d(t+"");return n(r,r),r}var e=s;return e.nodeType!==g?(e.before(e=d()),w(e)):N(e),n(e,e),e}function O(){if(i)return n(s,null),s;var t=document.createDocumentFragment(),r=document.createComment(""),e=d();return t.append(r,e),n(r,e),t}function P(t,r){if(i){var e=l;((e.f&E)===0||e.nodes.end===null)&&(e.nodes.end=s),y();return}t!==null&&t.before(r)}const L="5";typeof window<"u"&&((window.__svelte??={}).v??=new Set).add(L);export{P as a,n as b,O as c,b as f,C as t};

backend/static/export-viewer/_app/immutable/entry/start.YjNZv4co.js

@@ -1 +0,0 @@
-import{l as o,a as r}from"../chunks/eAGLaJx1.js";export{o as load_css,r as start};

backend/static/export-viewer/_app/immutable/nodes/0.DLesTSIe.js

@@ -1 +0,0 @@
-import{c as s,a as c}from"../chunks/RsTAN2PN.js";import{b as l,E as p,t as i}from"../chunks/CcONa1Mr.js";import{B as m}from"../chunks/BRDva_z9.js";function u(n,r,...e){var o=new m(n);l(()=>{const t=r()??null;o.ensure(t,t&&(a=>t(a,...e)))},p)}const f=!0,_=!1,g=Object.freeze(Object.defineProperty({__proto__:null,prerender:f,ssr:_},Symbol.toStringTag,{value:"Module"}));function h(n,r){var e=s(),o=i(e);u(o,()=>r.children),c(n,e)}export{h as component,g as universal};

backend/static/export-viewer/_app/immutable/nodes/1.D6Kk4lDA.js

@@ -1 +0,0 @@
-import{a as i,f as h}from"../chunks/RsTAN2PN.js";import{q as g,t as v,v as d,w as l,x as s,y as a,z as x}from"../chunks/CcONa1Mr.js";import{s as o}from"../chunks/Bb9JxzU7.js";import{s as _,p}from"../chunks/eAGLaJx1.js";const $={get error(){return p.error},get status(){return p.status}};_.updated.check;const m=$;var k=h("<h1> </h1> <p> </p>",1);function z(c,f){g(f,!0);var t=k(),r=v(t),n=s(r,!0);a(r);var e=x(r,2),u=s(e,!0);a(e),d(()=>{o(n,m.status),o(u,m.error?.message)}),i(c,t),l()}export{z as component};

backend/static/export-viewer/_app/version.json

@@ -1 +0,0 @@
-{"version":"1778876725548"}

backend/tests/common/mod.rs

@@ -255,3 +255,91 @@ pub async fn downloadable(pool: &PgPool, event_id: Uuid, export_type: &str) -> O
     .expect("downloadable")
     .flatten()
 }
+
+/// Insert an upload of `size` bytes, optionally already soft-deleted.
+pub async fn seed_upload(
+    pool: &PgPool,
+    event_id: Uuid,
+    user_id: Uuid,
+    size: i64,
+    deleted: bool,
+) -> Uuid {
+    sqlx::query_scalar(
+        "INSERT INTO upload (event_id, user_id, original_path, mime_type,
+                             original_size_bytes, deleted_at)
+         VALUES ($1, $2, 'originals/x.jpg', 'image/jpeg', $3,
+                 CASE WHEN $4 THEN NOW() ELSE NULL END)
+         RETURNING id",
+    )
+    .bind(event_id)
+    .bind(user_id)
+    .bind(size)
+    .bind(deleted)
+    .fetch_one(pool)
+    .await
+    .expect("seed upload")
+}
+
+/// Flip the moderation flags a ban sets.
+pub async fn set_user_moderation(pool: &PgPool, user_id: Uuid, banned: bool, hidden: bool) {
+    sqlx::query("UPDATE \"user\" SET is_banned = $2, uploads_hidden = $3 WHERE id = $1")
+        .bind(user_id)
+        .bind(banned)
+        .bind(hidden)
+        .execute(pool)
+        .await
+        .expect("set moderation");
+}
+
+/// SRC: `services/export.rs::query_uploads` — the visibility filter, verbatim, projected down to
+/// `(id, original_size_bytes)`. This is the row set that ACTUALLY lands in the archives.
+///
+/// Production builds this WHERE from `export_visibility_where!()`, shared with
+/// `estimate_export_bytes`. A copy here can pin the behaviour but CANNOT detect production moving
+/// away from it — that is what sharing the fragment is for, not this.
+pub async fn export_visible_uploads(pool: &PgPool, event_id: Uuid) -> Vec<(Uuid, i64)> {
+    sqlx::query_as(
+        "SELECT u.id, u.original_size_bytes
+         FROM upload u
+         JOIN \"user\" usr ON usr.id = u.user_id
+         WHERE u.event_id = $1 AND u.deleted_at IS NULL
+           AND usr.uploads_hidden = FALSE AND usr.is_banned = FALSE
+         GROUP BY u.id, usr.display_name
+         ORDER BY u.created_at ASC",
+    )
+    .bind(event_id)
+    .fetch_all(pool)
+    .await
+    .expect("export_visible_uploads")
+}
+
+/// SRC: `services/export.rs::estimate_export_bytes` — verbatim. Same caveat as above: production
+/// shares its WHERE with `query_uploads` via `export_visibility_where!()`, so these two copies
+/// agreeing proves the behaviour, not the absence of drift.
+pub async fn estimate_export_bytes(pool: &PgPool, event_id: Uuid) -> i64 {
+    let (bytes,): (i64,) = sqlx::query_as(
+        "SELECT COALESCE(SUM(u.original_size_bytes), 0)::bigint
+           FROM upload u
+           JOIN \"user\" usr ON usr.id = u.user_id
+          WHERE u.event_id = $1 AND u.deleted_at IS NULL
+            AND usr.uploads_hidden = FALSE AND usr.is_banned = FALSE",
+    )
+    .bind(event_id)
+    .fetch_one(pool)
+    .await
+    .expect("estimate_export_bytes");
+    bytes
+}
+
+/// SRC: `services/export.rs::ensure_export_space` — the armed-job count, verbatim.
+pub async fn armed_job_count(pool: &PgPool, event_id: Uuid) -> i64 {
+    let (n,): (i64,) = sqlx::query_as(
+        "SELECT COUNT(*) FROM export_job
+          WHERE event_id = $1 AND status IN ('pending', 'running')",
+    )
+    .bind(event_id)
+    .fetch_one(pool)
+    .await
+    .expect("armed_job_count");
+    n
+}

backend/tests/export_preflight.rs

@@ -0,0 +1,163 @@
+//! DB-backed tests for the export disk preflight.
+//!
+//! The keepsake used to be built with NO free-space check at all, and the failure that produced was
+//! not "the export failed" but "the deliverable is stuck and the escape hatch needs the space that
+//! isn't there":
+//!
+//!   1. A takedown bumps the epoch and re-arms both halves.
+//!   2. The ZIP hits ENOSPC partway through a multi-GB write.
+//!   3. The job row is now `failed` at the CURRENT epoch, so readiness
+//!      (`epoch = event.export_epoch AND status = 'done'`) is false and `GET /export/zip` 404s —
+//!      while the last good archive sits on disk, unreferenced and unreachable.
+//!   4. `POST /host/export/rebuild` re-arms the same doomed write.
+//!
+//! Two changes close it: reclaim the superseded generation BEFORE building (so peak usage is one
+//! generation, not two) and refuse up front with a number the host can act on.
+//!
+//! What these tests pin is the ESTIMATE — the part that decides. The arithmetic on top of it lives
+//! in `services/export.rs`'s unit tests; the filesystem selection lives in `is_superseded_archive`.
+//!
+//! ON DRIFT, precisely, because it is easy to overclaim here. The hazard is that `query_uploads`
+//! (which selects the rows the archives are built from) and `estimate_export_bytes` (which sizes
+//! them) could disagree — and an estimate missing rows the archive writes UNDER-reserves, the one
+//! direction that reintroduces the ENOSPC. **These tests cannot catch that**, and neither can any
+//! test in this harness: both sides here are `SRC:`-marked hand-copies in `tests/common/mod.rs`,
+//! so if production moved and the copies didn't, they would sit still and keep passing.
+//!
+//! That is fixed where it can be — the two queries now share one `export_visibility_where!()`
+//! fragment in `services/export.rs`, so they cannot diverge by construction. What is left for
+//! these tests is what the convention is genuinely good at: pinning the BEHAVIOUR, so a change
+//! that deliberately alters the filter has to come here and say so.
+
+mod common;
+
+use common::*;
+use sqlx::PgPool;
+
+/// The estimate must equal the sum over EXACTLY the rows `query_uploads` returns — computed from
+/// that row set, not from a restatement of its WHERE clause.
+///
+/// PINS: which uploads the preflight is allowed to count. Each excluded row below is excluded by a
+/// DIFFERENT predicate, so a change that drops or weakens any one of them fails here and has to be
+/// argued for. (It does not detect production drifting away from these copies — see the file
+/// header; `export_visibility_where!()` is what makes that impossible.)
+#[sqlx::test]
+async fn the_estimate_sums_exactly_the_rows_the_archive_will_contain(pool: PgPool) {
+    let event_id = seed_event(&pool, "wedding").await;
+
+    let visible = seed_user(&pool, event_id, "Anna").await;
+    let banned = seed_user(&pool, event_id, "Ben").await;
+    let hidden = seed_user(&pool, event_id, "Cara").await;
+
+    seed_upload(&pool, event_id, visible, 1_000, false).await;
+    seed_upload(&pool, event_id, visible, 2_500, false).await;
+    // Each of these is excluded from the archive by a DIFFERENT predicate.
+    seed_upload(&pool, event_id, visible, 9_000, true).await; // soft-deleted
+    seed_upload(&pool, event_id, banned, 9_000, false).await; // uploader banned
+    seed_upload(&pool, event_id, hidden, 9_000, false).await; // uploads hidden
+
+    set_user_moderation(&pool, banned, true, true).await;
+    set_user_moderation(&pool, hidden, false, true).await;
+
+    let rows = export_visible_uploads(&pool, event_id).await;
+    let expected: i64 = rows.iter().map(|(_, bytes)| bytes).sum();
+
+    assert_eq!(rows.len(), 2, "only Anna's two live uploads are archived");
+    assert_eq!(
+        estimate_export_bytes(&pool, event_id).await,
+        expected,
+        "the preflight must size the gallery the export will actually write"
+    );
+    assert_eq!(expected, 3_500);
+}
+
+/// An event with nothing to archive estimates zero rather than NULL.
+///
+/// PREVENTS: `SUM()` over no rows returning NULL and the decode blowing up — which would abort the
+/// export with a type error instead of building an (entirely legitimate) empty keepsake.
+#[sqlx::test]
+async fn an_empty_gallery_estimates_zero_not_null(pool: PgPool) {
+    let event_id = seed_event(&pool, "wedding").await;
+    assert_eq!(estimate_export_bytes(&pool, event_id).await, 0);
+
+    // And with a user who has uploaded nothing.
+    seed_user(&pool, event_id, "Anna").await;
+    assert_eq!(estimate_export_bytes(&pool, event_id).await, 0);
+}
+
+/// A release arms both halves, so the preflight sees a count of 2 and reserves for the pair.
+///
+/// PREVENTS: the concurrency under-reservation. `spawn_export_jobs` starts the ZIP and HTML workers
+/// at the same instant, and BOTH are gallery-sized (`Memories.zip` streams the original for every
+/// video and every image at or under 5 MB, all `Compression::Stored`). A worker reserving only for
+/// itself would see "it fits", its sibling would independently see the same, and together they
+/// would ENOSPC — which is why `required_free_bytes` multiplies by this count.
+#[sqlx::test]
+async fn a_release_arms_both_halves_so_the_preflight_reserves_for_two(pool: PgPool) {
+    let event_id = seed_event(&pool, "wedding").await;
+    let user = seed_user(&pool, event_id, "Anna").await;
+    seed_upload(&pool, event_id, user, 1_000, false).await;
+
+    assert_eq!(
+        armed_job_count(&pool, event_id).await,
+        0,
+        "nothing is armed before the release"
+    );
+
+    let epoch = release_gallery(&pool, "wedding").await.expect("released");
+    assert_eq!(
+        armed_job_count(&pool, event_id).await,
+        2,
+        "a release arms zip AND html — both compete for the same disk"
+    );
+
+    // A worker that has claimed its half is still competing; `running` must keep counting.
+    assert!(claim_job(&pool, event_id, "zip", epoch).await);
+    assert_eq!(
+        armed_job_count(&pool, event_id).await,
+        2,
+        "claiming moves pending -> running, which must not drop out of the reservation"
+    );
+
+    // Only a FINISHED half stops competing.
+    assert!(finalize_job(&pool, event_id, "zip", epoch, "exports/Gallery.zip").await);
+    assert_eq!(
+        armed_job_count(&pool, event_id).await,
+        1,
+        "a done half no longer needs space reserved for it"
+    );
+}
+
+/// A ViewerOnly regeneration re-arms only the HTML half, so the preflight reserves for one.
+///
+/// PREVENTS: over-reservation refusing a rebuild that fits perfectly well. Moderating a comment
+/// carries the finished ZIP forward untouched; demanding room for a second copy of it would fail
+/// the one operation that needs no new gallery-sized write at all.
+#[sqlx::test]
+async fn a_viewer_only_regeneration_reserves_for_one_half(pool: PgPool) {
+    let event_id = seed_event(&pool, "wedding").await;
+    let user = seed_user(&pool, event_id, "Anna").await;
+    seed_upload(&pool, event_id, user, 1_000, false).await;
+
+    let epoch = release_gallery(&pool, "wedding").await.expect("released");
+    for t in ["zip", "html"] {
+        assert!(claim_job(&pool, event_id, t, epoch).await);
+        assert!(finalize_job(&pool, event_id, t, epoch, &format!("exports/{t}")).await);
+    }
+    assert_eq!(armed_job_count(&pool, event_id).await, 0);
+
+    // A moderated comment: bump the epoch, carry the ZIP forward, re-arm only the viewer.
+    let (_, _, next) = bump_epoch(&pool, "wedding").await.expect("bumped");
+    assert!(
+        carry_zip_forward(&pool, event_id, next).await,
+        "the finished ZIP is re-stamped, not rebuilt"
+    );
+    let mut conn = pool.acquire().await.expect("acquire");
+    enqueue_types_at_epoch(&mut conn, event_id, next, &["html"]).await;
+
+    assert_eq!(
+        armed_job_count(&pool, event_id).await,
+        1,
+        "only the viewer is being rebuilt, so only one archive's worth of space is needed"
+    );
+}

backend/tests/failed_original_sweep.rs

@@ -0,0 +1,352 @@
+//! DB-backed tests for the deleted-media sweep (`services/maintenance.rs`).
+//!
+//! Context, in two halves.
+//!
+//! The compression worker deliberately no longer deletes an upload's original when its transcode
+//! fails — a transient ENOSPC or a codec panic must never destroy the only copy of a photo a guest
+//! cannot retake. But the row is soft-deleted and the uploader's quota IS refunded, so those bytes
+//! become invisible, unowned and free.
+//!
+//! The SAME hole was reachable by the ordinary path, and that one is not an edge case at all:
+//! `soft_delete_in_event` refunds `total_upload_bytes` on every guest or host delete and nothing
+//! removed the files, so the quota stopped bounding the disk. Upload 500 MB, delete, quota back to
+//! zero, upload another 500 MB — a guest curating their camera roll, which is what people do. The
+//! sweep used to reach only `compression_status = 'failed'`, so it never touched this case; the
+//! test below that now asserts an owner-deleted upload IS reclaimed is the one that used to assert
+//! the opposite.
+//!
+//! Two windows, because the two deletes mean different things: 14 days for a failure an operator
+//! may want to investigate, 24 hours for a removal someone asked for (14 days outlives the whole
+//! event, so a deliberate delete would never reclaim anything while it mattered).
+//!
+//! The selection predicate is the whole safety argument — it must reach both leftovers and never a
+//! live upload — so that is what these pin, following the same "reproduce the SQL verbatim" pattern
+//! as `upload_concurrency.rs`. `#[sqlx::test]` gives each test a fresh, migrated database.
+
+mod common;
+
+use common::*;
+use sqlx::PgPool;
+use uuid::Uuid;
+
+const FAILED_DAYS: i64 = 14;
+const DELETED_HOURS: i64 = 24;
+
+/// SRC: `services/maintenance.rs::cleanup_deleted_media` — the selection, verbatim.
+async fn sweep_selects(pool: &PgPool, failed_days: i64, deleted_hours: i64) -> Vec<Uuid> {
+    type Row = (Uuid, String, Option<String>, Option<String>, Option<String>);
+    sqlx::query_as::<_, Row>(
+        "SELECT id, original_path, preview_path, display_path, thumbnail_path FROM upload
+         WHERE deleted_at IS NOT NULL
+           AND CASE WHEN compression_status = 'failed'
+                    THEN deleted_at < NOW() - ($1 || ' days')::interval
+                    ELSE deleted_at < NOW() - ($2 || ' hours')::interval
+               END
+           AND (original_path <> '' OR preview_path IS NOT NULL
+                OR display_path IS NOT NULL OR thumbnail_path IS NOT NULL)",
+    )
+    .bind(failed_days.to_string())
+    .bind(deleted_hours.to_string())
+    .fetch_all(pool)
+    .await
+    .expect("sweep query")
+    .into_iter()
+    .map(|(id, ..)| id)
+    .collect()
+}
+
+/// Seed an upload aged `deleted_hours_ago` (None = live), with optional derivative paths.
+async fn seed_aged_upload(
+    pool: &PgPool,
+    event_id: Uuid,
+    user_id: Uuid,
+    status: &str,
+    deleted_hours_ago: Option<i64>,
+    original_path: &str,
+    derivatives: bool,
+) -> Uuid {
+    sqlx::query_scalar(
+        "INSERT INTO upload (event_id, user_id, original_path, mime_type, original_size_bytes,
+                             compression_status, deleted_at,
+                             preview_path, display_path, thumbnail_path)
+         VALUES ($1, $2, $3, 'image/jpeg', 1000, $4,
+                 CASE WHEN $5::bigint IS NULL THEN NULL
+                      ELSE NOW() - ($5::text || ' hours')::interval END,
+                 CASE WHEN $6 THEN 'previews/p.jpg' END,
+                 CASE WHEN $6 THEN 'displays/d.jpg' END,
+                 CASE WHEN $6 THEN 'thumbs/t.jpg' END)
+         RETURNING id",
+    )
+    .bind(event_id)
+    .bind(user_id)
+    .bind(original_path)
+    .bind(status)
+    .bind(deleted_hours_ago)
+    .bind(derivatives)
+    .fetch_one(pool)
+    .await
+    .expect("seed upload")
+}
+
+/// A live upload is untouchable no matter how the windows are configured.
+///
+/// PREVENTS: the catastrophic loosening. Everything else here is about reclaiming more; this is the
+/// one assertion that must never bend.
+#[sqlx::test]
+async fn a_live_upload_is_never_selected(pool: PgPool) {
+    let event_id = seed_event(&pool, "sweep-live").await;
+    let user_id = seed_user(&pool, event_id, "Sweeper").await;
+
+    for status in ["done", "failed", "processing", "pending"] {
+        let live = seed_aged_upload(
+            &pool,
+            event_id,
+            user_id,
+            status,
+            None,
+            "originals/e/live.jpg",
+            true,
+        )
+        .await;
+        assert!(
+            !sweep_selects(&pool, FAILED_DAYS, DELETED_HOURS)
+                .await
+                .contains(&live),
+            "a non-deleted upload with status {status} must never be swept"
+        );
+    }
+}
+
+/// THE FIX. An upload a guest or host deliberately deleted is reclaimed once past 24 hours.
+///
+/// PREVENTS: the regression back to a sweep scoped to `compression_status = 'failed'`, which is
+/// what let the quota stop bounding the disk. This assertion is the inverse of the one this file
+/// used to make.
+#[sqlx::test]
+async fn a_deliberately_deleted_upload_is_reclaimed_after_a_day(pool: PgPool) {
+    let event_id = seed_event(&pool, "sweep-deleted").await;
+    let user_id = seed_user(&pool, event_id, "Curator").await;
+
+    let deleted = seed_aged_upload(
+        &pool,
+        event_id,
+        user_id,
+        "done",
+        Some(48),
+        "originals/e/owner.jpg",
+        true,
+    )
+    .await;
+    // Still inside the window — a mis-tap is recoverable for a day.
+    let recent = seed_aged_upload(
+        &pool,
+        event_id,
+        user_id,
+        "done",
+        Some(2),
+        "originals/e/recent.jpg",
+        true,
+    )
+    .await;
+
+    let selected = sweep_selects(&pool, FAILED_DAYS, DELETED_HOURS).await;
+    assert!(
+        selected.contains(&deleted),
+        "a deliberate delete past the window must be reclaimed — this is the leak"
+    );
+    assert!(
+        !selected.contains(&recent),
+        "a delete inside the window keeps its recovery grace"
+    );
+}
+
+/// The two windows are independent: a failure is retained far longer than a deliberate delete.
+///
+/// PREVENTS: collapsing them into one. Applying 24h to failures would destroy the recovery window
+/// the retained-original fix exists to provide; applying 14 days to deliberate deletes would mean
+/// nothing is ever reclaimed during an event.
+#[sqlx::test]
+async fn the_two_retention_windows_do_not_bleed_into_each_other(pool: PgPool) {
+    let event_id = seed_event(&pool, "sweep-windows").await;
+    let user_id = seed_user(&pool, event_id, "Windows").await;
+
+    // 48h old: past the deliberate window, nowhere near the failure window.
+    let failed_recent = seed_aged_upload(
+        &pool,
+        event_id,
+        user_id,
+        "failed",
+        Some(48),
+        "originals/e/f-recent.jpg",
+        false,
+    )
+    .await;
+    let deleted_same_age = seed_aged_upload(
+        &pool,
+        event_id,
+        user_id,
+        "done",
+        Some(48),
+        "originals/e/d-same.jpg",
+        false,
+    )
+    .await;
+    // 30 days old: past both.
+    let failed_old = seed_aged_upload(
+        &pool,
+        event_id,
+        user_id,
+        "failed",
+        Some(30 * 24),
+        "originals/e/f-old.jpg",
+        false,
+    )
+    .await;
+
+    let selected = sweep_selects(&pool, FAILED_DAYS, DELETED_HOURS).await;
+    assert!(
+        !selected.contains(&failed_recent),
+        "a 2-day-old compression failure is still inside its 14-day recovery window"
+    );
+    assert!(
+        selected.contains(&deleted_same_age),
+        "a deliberate delete of the same age is past its 24-hour window"
+    );
+    assert!(
+        selected.contains(&failed_old),
+        "a 30-day-old failure is past both windows"
+    );
+}
+
+/// Boundary behaviour on both windows.
+#[sqlx::test]
+async fn retention_windows_are_honoured_at_the_boundary(pool: PgPool) {
+    let event_id = seed_event(&pool, "sweep-boundary").await;
+    let user_id = seed_user(&pool, event_id, "Boundary").await;
+
+    let cases = [
+        ("failed", 13 * 24, false, "13 days"),
+        ("failed", 15 * 24, true, "15 days"),
+        ("done", 23, false, "23 hours"),
+        ("done", 25, true, "25 hours"),
+    ];
+    for (status, hours, expected, label) in cases {
+        let id = seed_aged_upload(
+            &pool,
+            event_id,
+            user_id,
+            status,
+            Some(hours),
+            "originals/e/b.jpg",
+            false,
+        )
+        .await;
+        assert_eq!(
+            sweep_selects(&pool, FAILED_DAYS, DELETED_HOURS)
+                .await
+                .contains(&id),
+            expected,
+            "a {status} upload deleted {label} ago: expected swept={expected}"
+        );
+        sqlx::query("DELETE FROM upload WHERE id = $1")
+            .bind(id)
+            .execute(&pool)
+            .await
+            .expect("clean up");
+    }
+}
+
+/// A row is re-selected until EVERY one of its paths is cleared.
+///
+/// PREVENTS: two failures at once. The sweep used to clear `original_path` alone, which was right
+/// for its only case (a failed compression produces no derivatives) but leaves preview, display and
+/// thumbnail on disk the moment it reaches a successfully processed upload — three files per
+/// upload, none of them counted in `original_size_bytes`, that nothing else ever removes. And a row
+/// whose paths are all cleared must stop coming back, or every hourly tick logs a phantom reclaim
+/// forever.
+#[sqlx::test]
+async fn a_row_is_reselected_until_every_path_is_cleared(pool: PgPool) {
+    let event_id = seed_event(&pool, "sweep-idempotent").await;
+    let user_id = seed_user(&pool, event_id, "Idem").await;
+    let id = seed_aged_upload(
+        &pool,
+        event_id,
+        user_id,
+        "done",
+        Some(48),
+        "originals/e/once.jpg",
+        true,
+    )
+    .await;
+
+    assert_eq!(sweep_selects(&pool, FAILED_DAYS, DELETED_HOURS).await, [id]);
+
+    // Clearing only the original is NOT enough — the derivatives are still on disk.
+    sqlx::query("UPDATE upload SET original_path = '' WHERE id = $1")
+        .bind(id)
+        .execute(&pool)
+        .await
+        .expect("clear original");
+    assert_eq!(
+        sweep_selects(&pool, FAILED_DAYS, DELETED_HOURS).await,
+        [id],
+        "derivatives left behind must keep the row selected"
+    );
+
+    sqlx::query(
+        "UPDATE upload SET preview_path = NULL, display_path = NULL, thumbnail_path = NULL
+         WHERE id = $1",
+    )
+    .bind(id)
+    .execute(&pool)
+    .await
+    .expect("clear derivatives");
+    assert!(
+        sweep_selects(&pool, FAILED_DAYS, DELETED_HOURS)
+            .await
+            .is_empty(),
+        "a fully swept row must not come back"
+    );
+}
+
+/// The derivative backfill must never resurrect what the sweep just reclaimed.
+///
+/// PREVENTS: an interaction, not a bug in either piece. The sweep nulls `preview_path`, and
+/// `backfill_stale_derivatives` selects on `display_path IS NULL AND preview_path IS NOT NULL` —
+/// close enough that a future edit to either could have the backfill re-decode an original that is
+/// no longer on disk, on every boot. `deleted_at IS NULL` is what keeps them apart.
+#[sqlx::test]
+async fn the_backfill_ignores_swept_rows(pool: PgPool) {
+    let event_id = seed_event(&pool, "sweep-backfill").await;
+    let user_id = seed_user(&pool, event_id, "Backfill").await;
+    seed_aged_upload(
+        &pool,
+        event_id,
+        user_id,
+        "done",
+        Some(48),
+        "originals/e/gone.jpg",
+        true,
+    )
+    .await;
+
+    // SRC: `services/compression.rs::backfill_stale_derivatives` — the selection, verbatim.
+    let backfilled: Vec<(Uuid, String, String)> = sqlx::query_as(
+        "SELECT id, original_path, mime_type FROM upload
+             WHERE deleted_at IS NULL AND mime_type LIKE 'image/%'
+               AND original_path IS NOT NULL
+               AND (
+                     (display_path IS NULL AND preview_path IS NOT NULL)
+                  OR derivatives_rev < $1
+               )",
+    )
+    .bind(1i16)
+    .fetch_all(&pool)
+    .await
+    .expect("backfill query");
+
+    assert!(
+        backfilled.is_empty(),
+        "a soft-deleted row must be invisible to the backfill, before or after sweeping"
+    );
+}

backend/tests/upload_idempotency.rs

@@ -0,0 +1,175 @@
+//! DB-backed tests for the upload idempotency key (migration 022).
+//!
+//! The guarantee under test is the one thing standing between a lost response and a duplicated
+//! wedding photo: a retry of an upload that already committed must NOT create a second row, and
+//! must not charge the guest's storage quota twice. The whole mechanism is SQL — a partial unique
+//! index plus `ON CONFLICT DO NOTHING` — so it is tested against a real database with the real
+//! migrations applied, using the same statements `src/` runs.
+
+mod common;
+
+use common::*;
+use sqlx::PgPool;
+use uuid::Uuid;
+
+/// SRC: `models/upload.rs::Upload::create` — the insert, verbatim.
+///
+/// Returns the new row's id, or `None` when the key was already stored. The handler treats
+/// `None` as "a concurrent retry won" and replays the stored row instead of committing.
+async fn create_upload(
+    pool: &PgPool,
+    event_id: Uuid,
+    user_id: Uuid,
+    original_path: &str,
+    client_upload_id: Option<Uuid>,
+) -> Option<Uuid> {
+    let row: Option<(Uuid,)> = sqlx::query_as(
+        "INSERT INTO upload (event_id, user_id, original_path, mime_type, original_size_bytes, caption, client_upload_id)
+         VALUES ($1, $2, $3, $4, $5, $6, $7)
+         ON CONFLICT (client_upload_id) WHERE client_upload_id IS NOT NULL DO NOTHING
+         RETURNING id",
+    )
+    .bind(event_id)
+    .bind(user_id)
+    .bind(original_path)
+    .bind("image/jpeg")
+    .bind(1_000i64)
+    .bind(Option::<String>::None)
+    .bind(client_upload_id)
+    .fetch_optional(pool)
+    .await
+    .expect("create_upload");
+    row.map(|(id,)| id)
+}
+
+/// SRC: `models/upload.rs::Upload::find_by_client_upload_id` — the lookup, verbatim.
+async fn find_by_key(pool: &PgPool, user_id: Uuid, client_upload_id: Uuid) -> Option<Uuid> {
+    let row: Option<(Uuid,)> = sqlx::query_as(
+        "SELECT id FROM upload
+         WHERE client_upload_id = $1 AND user_id = $2 AND deleted_at IS NULL",
+    )
+    .bind(client_upload_id)
+    .bind(user_id)
+    .fetch_optional(pool)
+    .await
+    .expect("find_by_key");
+    row.map(|(id,)| id)
+}
+
+async fn upload_count(pool: &PgPool) -> i64 {
+    sqlx::query_scalar("SELECT COUNT(*) FROM upload")
+        .fetch_one(pool)
+        .await
+        .expect("upload_count")
+}
+
+/// The core guarantee. A phone that loses the response and re-sends the same photo gets the
+/// original row back, not a second copy in the gallery and a second charge against its quota.
+#[sqlx::test]
+async fn the_same_key_can_only_ever_store_one_upload(pool: PgPool) {
+    let event_id = seed_event(&pool, "wedding").await;
+    let user_id = seed_user(&pool, event_id, "Wackelige Wanda").await;
+    let key = Uuid::new_v4();
+
+    let first = create_upload(&pool, event_id, user_id, "originals/a.jpg", Some(key)).await;
+    assert!(first.is_some(), "the first attempt must store the upload");
+
+    // The retry: same key, and (as after a real re-send) a different file on disk.
+    let second = create_upload(&pool, event_id, user_id, "originals/b.jpg", Some(key)).await;
+    assert!(
+        second.is_none(),
+        "a retry of a committed upload must not insert a second row"
+    );
+    assert_eq!(upload_count(&pool).await, 1);
+
+    // And the handler can find the winner to replay it.
+    assert_eq!(find_by_key(&pool, user_id, key).await, first);
+}
+
+/// The index must not over-reach. Two genuinely different photos carry different keys and must
+/// both land — this is the ordinary case, and breaking it would silently drop uploads.
+#[sqlx::test]
+async fn different_keys_are_different_uploads(pool: PgPool) {
+    let event_id = seed_event(&pool, "wedding").await;
+    let user_id = seed_user(&pool, event_id, "Fleißige Frieda").await;
+
+    for _ in 0..5 {
+        assert!(
+            create_upload(
+                &pool,
+                event_id,
+                user_id,
+                "originals/x.jpg",
+                Some(Uuid::new_v4())
+            )
+            .await
+            .is_some()
+        );
+    }
+    assert_eq!(upload_count(&pool).await, 5);
+}
+
+/// The index is PARTIAL, and this is why. Every upload that predates migration 022, and any
+/// client that doesn't send a key, carries NULL — if those collided, the first such upload would
+/// block every subsequent one and the whole event would fail after one photo.
+#[sqlx::test]
+async fn uploads_without_a_key_never_collide(pool: PgPool) {
+    let event_id = seed_event(&pool, "wedding").await;
+    let user_id = seed_user(&pool, event_id, "Alte Anna").await;
+
+    for _ in 0..5 {
+        assert!(
+            create_upload(&pool, event_id, user_id, "originals/legacy.jpg", None)
+                .await
+                .is_some(),
+            "a NULL key must never be treated as a duplicate"
+        );
+    }
+    assert_eq!(upload_count(&pool).await, 5);
+}
+
+/// The lookup is scoped to the owner. The key alone is unique, so this can only matter if a key
+/// ever repeated across users — but a replay that handed one guest another guest's upload row
+/// would be a data leak, so the scope is asserted rather than assumed.
+#[sqlx::test]
+async fn the_replay_lookup_never_crosses_users(pool: PgPool) {
+    let event_id = seed_event(&pool, "wedding").await;
+    let owner = seed_user(&pool, event_id, "Besitzerin Bea").await;
+    let other = seed_user(&pool, event_id, "Fremder Franz").await;
+    let key = Uuid::new_v4();
+
+    let id = create_upload(&pool, event_id, owner, "originals/a.jpg", Some(key)).await;
+
+    assert_eq!(find_by_key(&pool, owner, key).await, id);
+    assert_eq!(
+        find_by_key(&pool, other, key).await,
+        None,
+        "another guest's retry must not resolve to this upload"
+    );
+}
+
+/// A deleted photo must not be resurrected by a stale queue item. If the guest uploaded, deleted,
+/// and their queue then retried the original request, replaying the deleted row would put the
+/// photo they removed back in the gallery — so the lookup excludes soft-deleted rows and the
+/// retry becomes a fresh upload instead.
+#[sqlx::test]
+async fn a_deleted_upload_is_not_replayed(pool: PgPool) {
+    let event_id = seed_event(&pool, "wedding").await;
+    let user_id = seed_user(&pool, event_id, "Reumütige Rita").await;
+    let key = Uuid::new_v4();
+
+    let id = create_upload(&pool, event_id, user_id, "originals/a.jpg", Some(key))
+        .await
+        .expect("first insert");
+    sqlx::query("UPDATE upload SET deleted_at = NOW() WHERE id = $1")
+        .bind(id)
+        .execute(&pool)
+        .await
+        .expect("soft delete");
+
+    assert_eq!(
+        find_by_key(&pool, user_id, key).await,
+        None,
+        "a soft-deleted upload must not be replayed"
+    );
+}

docker-compose.build.yml

@@ -0,0 +1,24 @@
+# Build overlay. NOT loaded automatically (same convention as docker-compose.dev.yml — this
+# repo deliberately never uses the auto-loaded `docker-compose.override.yml` name).
+#
+# Used ONLY on a workstation that builds and pushes images. Never on the production server,
+# which pulls: see the note on the `app` service in docker-compose.yml.
+#
+# Restores the `build:` keys that production omits, so `docker compose build` still works from
+# a single source of truth for the build context paths:
+#
+#   docker compose -f docker-compose.yml -f docker-compose.build.yml build
+#
+# For the actual release build use buildx directly instead — it is what produces linux/amd64
+# images from an arm64 Mac and pushes them in one step (see DEPLOYMENT_RUNBOOK.md §6):
+#
+#   docker buildx build --platform linux/amd64 -t registry.mc02.dev/eventsnap/app:vX.Y.Z --push ./backend
+services:
+  app:
+    build:
+      context: ./backend
+      dockerfile: Dockerfile
+  frontend:
+    build:
+      context: ./frontend
+      dockerfile: Dockerfile

docker-compose.dev.yml

@@ -11,3 +11,36 @@ services:
     # is tolerated (warned) rather than rejected.
     environment:
       APP_ENV: development
+      # `.env` sets DATABASE_URL to @localhost for the run-backend-natively workflow
+      # (the db port is published above for that). When the app runs IN a container,
+      # localhost is the app itself — point it at the `db` service instead. Creds are
+      # interpolated from .env so nothing is hardcoded.
+      DATABASE_URL: postgres://${POSTGRES_USER}:${POSTGRES_PASSWORD}@db:5432/${POSTGRES_DB}
+      # `.env` sets MEDIA_PATH to a HOST path (/home/fabi/EventSnap/media) for the
+      # run-backend-natively workflow. In a container that path doesn't exist and the
+      # app user can't create it from `/`, so every upload 500s with EACCES. The media
+      # volume is mounted at /media (see docker-compose.yml) — point the app there.
+      MEDIA_PATH: /media
+      # Set here purely so local dev needs no `.env` at all. The `$` are doubled because
+      # THIS is a `docker-compose.yml` `environment:` value, where Compose does interpolate.
+      #
+      # DO NOT COPY THE DOUBLING INTO `.env`. An earlier version of this comment claimed
+      # production had "the same bug" and told you to escape the hash as `$$` there — that is
+      # wrong and it breaks a working deployment. Compose uses SINGLE-QUOTED `env_file` values
+      # literally, which is the form `.env.example` ships, so the `$` segments survive intact;
+      # doubling them produces a 74-character string that `looks_bcrypt` rejects and the app
+      # refuses to boot on. Verified with `docker compose exec app printenv`.
+      ADMIN_PASSWORD_HASH: "$$2b$$12$$PAteqCNpsbm6d0HTJcywfOaUovjAU.iNVlsL7EDYaRC/z4P/xv7ye"
+      # Smoke-testing the comment kill-switch: boot-time flag, so it needs a restart
+      # (not an admin-UI toggle). Backend rejects new comments (403) and the frontend
+      # hides the whole comment UI. Flip back to true (or drop this line) to restore.
+      COMMENTS_ENABLED: "false"
+
+  caddy:
+    # The caddy service has no env_file, so the Caddyfile's `{$DOMAIN}` would expand to empty
+    # and the site block would collapse into a malformed global block. Supply it for local dev
+    # (from .env → localhost, which Caddy serves with a local self-signed cert). Production
+    # already wires DOMAIN into caddy's `environment:`, guarded with `:?` so an unset value
+    # fails the command instead of silently producing a site with no address.
+    environment:
+      DOMAIN: ${DOMAIN}

docker-compose.yml

@@ -1,7 +1,21 @@
+# Docker's default json-file driver has NO rotation at all, and every container writes to the
+# same filesystem as postgres_data, media_data and exports_data. Filling that filesystem does
+# not degrade one subsystem — Postgres stops being able to write and the whole event goes down
+# (see README "Sizing the disk"). This caps logs at 30 MB per service, permanently.
+#
+# Paired with RUST_LOG in .env: without it the app falls back to `eventsnap_backend=debug,
+# tower_http=debug` (main.rs), which is a debug line per HTTP request including every preview.
+x-logging: &default-logging
+  driver: json-file
+  options:
+    max-size: "10m"
+    max-file: "3"
+
 services:
   db:
     image: postgres:16-alpine
     restart: unless-stopped
+    logging: *default-logging
     env_file: .env
     environment:
       POSTGRES_USER: ${POSTGRES_USER}
@@ -17,21 +31,74 @@ services:
     deploy:
       resources:
         limits:
-          memory: 512M
+          # 1G, not 512M. Postgres 16's default shared_buffers plus a pool of backends
+          # leaves very little headroom at 512M, and an OOM here does not degrade one
+          # feature — it takes the event down, because every request path touches the
+          # database.
+          #
+          # `.env.example` now sets DATABASE_MAX_CONNECTIONS=15, sized to the 2 vCPU this
+          # box has rather than to the guest count: since migration 024 replaced the feed
+          # view's GROUP BY with scalar subqueries, a feed page costs well under a
+          # millisecond, so connections are no longer spent waiting. Raising it back
+          # toward 30 means raising this limit with it.
+          memory: 1G
 
   app:
-    build:
-      context: ./backend
-      dockerfile: Dockerfile
+    # Production PULLS a prebuilt image; it never compiles. A release build of this crate is
+    # fat-LTO over 427 dependencies (see backend/Cargo.toml [profile.release]) and peaks well
+    # above the RAM a 4 GB box has spare with the stack running — and a rollback would be a
+    # second build under pressure. Images are built on a workstation and pushed; see
+    # docker-compose.build.yml and DEPLOYMENT_RUNBOOK.md.
+    #
+    # There is deliberately NO `build:` key here: without one, a wrong tag fails instantly with
+    # "manifest unknown" instead of silently starting a 45-minute compile on the event server.
+    # The `:?` form fails loudly on an unset variable rather than resolving to an empty tag.
+    image: registry.mc02.dev/eventsnap/app:${EVENTSNAP_VERSION:?set EVENTSNAP_VERSION in .env}
     restart: unless-stopped
+    logging: *default-logging
     env_file: .env
     environment:
+      # Default to info. Without this the code fallback in main.rs applies, which is
+      # `eventsnap_backend=debug,tower_http=debug` — a line per HTTP request AND per
+      # response, including every preview fetch, for a multi-day run. The x-logging cap
+      # above bounds the disk cost but not the CPU/IO one.
+      #
+      # Set here rather than only in `.env` because a stock deploy sets RUST_LOG nowhere,
+      # and this is the layer an operator will actually find when they need to raise it
+      # for a single event (`RUST_LOG=eventsnap_backend=debug docker compose up -d app`).
+      RUST_LOG: ${RUST_LOG:-info}
       # Activates the production secret guard in config.rs — refuses to boot with
       # placeholder JWT_SECRET / ADMIN_PASSWORD_HASH.
       APP_ENV: production
+      # Pinned beside MEDIA_PATH for the same reason, and because nothing validates it:
+      # config.rs defaults it to /exports but never checks that it is a mount or that it
+      # differs from media_path. A stray EXPORT_PATH in .env builds the keepsake into the
+      # container's writable layer, where it passes every health check and disk preflight
+      # and then evaporates on the next `up -d`.
+      EXPORT_PATH: /exports
+      # The media volume is mounted at /media (below), so the app MUST write there.
+      # Pin it here rather than trusting .env: if MEDIA_PATH in .env points elsewhere
+      # (e.g. a host path used for running the backend natively) the container can't
+      # create it and every upload 500s with EACCES. `environment` overrides `env_file`,
+      # so this is authoritative for the container.
+      MEDIA_PATH: /media
+      # Pinned for the same reason as MEDIA_PATH: `environment` beats `env_file`, so this cannot
+      # be lost by an operator who copies `.env.example` and edits only the secrets — which is
+      # the likely path, and `.env.example` ships the generic default of `true`.
+      #
+      # This is a product decision for this event, not a technical one: guests should be present
+      # at the party, not in a comment thread. Likes and captions stay on and are unaffected.
+      # Boot-time only, so changing it means `docker compose up -d`, not an admin toggle.
+      # To re-enable comments, delete this line and set COMMENTS_ENABLED in .env.
+      COMMENTS_ENABLED: "false"
     depends_on:
       db:
         condition: service_healthy
+    # Longer than the app's own 10s shutdown backstop (main.rs SHUTDOWN_GRACE), because Docker's
+    # default stop timeout is ALSO 10s — so a redeploy raced the graceful drain and could SIGKILL
+    # the process at the exact moment it was finishing, truncating the in-flight upload the
+    # graceful shutdown exists to protect. The app always exits well before 20s.
+    stop_grace_period: 20s
     volumes:
       - media_data:/media
       # Export archives live OUTSIDE /media so the public media ServeDir can't
@@ -40,7 +107,11 @@ services:
     expose:
       - "3000"
     healthcheck:
-      test: ["CMD-SHELL", "wget -q -O- http://localhost:3000/health || exit 1"]
+      # Use 127.0.0.1, NOT localhost: the app binds IPv4 (0.0.0.0) but `localhost`
+      # resolves to ::1 (IPv6) first inside the container, so a localhost probe gets
+      # "connection refused" and the container never turns healthy — which would leave
+      # Caddy (gated on `condition: service_healthy` below) blocked forever on boot.
+      test: ["CMD-SHELL", "wget -q -O- http://127.0.0.1:3000/health || exit 1"]
       interval: 10s
       timeout: 5s
       retries: 5
@@ -53,21 +124,35 @@ services:
           memory: 1G
 
   frontend:
-    build:
-      context: ./frontend
-      dockerfile: Dockerfile
+    # Pulled, not built — see the note on `app` above.
+    image: registry.mc02.dev/eventsnap/frontend:${EVENTSNAP_VERSION:?set EVENTSNAP_VERSION in .env}
     restart: unless-stopped
+    logging: *default-logging
     env_file: .env
     environment:
       # adapter-node behind Caddy TLS needs the public origin for CSRF checks on
       # POST form actions — without it they fail only in production.
-      ORIGIN: "https://${DOMAIN}"
+      # `:?` for the same reason EVENTSNAP_VERSION uses it. A blank DOMAIN doesn't fail — it
+      # produces `https://` here and collapses the Caddyfile's site block below, so the stack
+      # comes up with no TLS and no site and the only symptom is a browser error.
+      ORIGIN: "https://${DOMAIN:?set DOMAIN in .env}"
+      # V8 sizes its old-space heap from the cgroup limit, but lands on ~101% of it (measured:
+      # heap_size_limit 259 MB inside a 256M container). So the JS heap ceiling sits ABOVE the
+      # container's entire budget — before base RSS (~60-90 MB), the C++ heap, or SSR response
+      # buffers, which are external memory V8 doesn't count at all. The practical effect is that
+      # V8 can never reach its own limit and run an emergency GC, so the only backpressure is a
+      # kernel SIGKILL: an arrival burst of ~100 guests SSR-rendering /join and /feed OOM-kills
+      # node, guests get the "Wir sind gleich zurück" page, it restarts, and the burst is still
+      # there. Setting the ceiling below the cgroup limit restores GC as the first line of defence.
+      NODE_OPTIONS: "--max-old-space-size=160"
     depends_on:
       - app
     expose:
       - "3001"
     healthcheck:
-      test: ["CMD-SHELL", "wget -q -O- http://localhost:3001/ >/dev/null 2>&1 || exit 1"]
+      # 127.0.0.1, not localhost — see the app healthcheck note above (IPv4 bind vs
+      # ::1 resolution would leave this container permanently unhealthy).
+      test: ["CMD-SHELL", "wget -q -O- http://127.0.0.1:3001/ >/dev/null 2>&1 || exit 1"]
       interval: 10s
       timeout: 5s
       retries: 5
@@ -80,6 +165,13 @@ services:
   caddy:
     image: caddy:2-alpine
     restart: unless-stopped
+    logging: *default-logging
+    environment:
+      # The Caddyfile's site address is `{$DOMAIN}`, read from THIS container's env.
+      # Without it, `{$DOMAIN}` expands to empty, the site block collapses, and Caddy
+      # serves nothing / fails to obtain a TLS cert. `env_file` alone wouldn't help —
+      # Caddy needs it in `environment`, and this keeps the Caddyfile the single source.
+      DOMAIN: ${DOMAIN:?set DOMAIN in .env}
     ports:
       - "80:80"
       - "443:443"

docs/FEATURES.md

@@ -14,12 +14,12 @@ Status legend: **✓ shipped** · **◐ partial** · **◯ planned** · **✗ ou
 |---------------------------------------------------------|:-----:|:-----:|:-----:|-----------------------------------------------------------------------|
 | **Onboarding & sessions**                               |       |       |       |                                                                       |
 | Join via shared event link / QR code                    |   ✓   |   ✓   |   ✓   | Name-only registration; server issues JWT + 4-digit PIN               |
-| First-visit guided tour (4 steps)                       |   ✓   |   ✓   |   ✓   | Dismissed once, flag in `localStorage`                                |
+| First-visit guided tour (6 steps)                       |   ✓   |   ✓   |   ✓   | Welcome, upload, hashtags, long-press, hell/dunkel, PIN (+ privacy-note pointer when one is set). Dismissed once, flag in `localStorage` |
 | Persistent 30-day session                               |   ✓   |   ✓   |   ✓   | JWT in `localStorage`; refreshed on activity                          |
 | Sign in on another device using name + PIN              |   ✓   |   ✓   |   ✓   | 3 wrong PINs → 15-min lockout                                         |
 | "Ich habe bereits einen Account" link on the join page  |   ✓   |   ✓   |   ✓   | Small inline link → `/recover` (name + PIN)                           |
 | View / copy own PIN any time ("My Account")             |   ✓   |   ✓   |   ✓   | Read from `localStorage`; never sent back from the server             |
-| Log out / "Leave event"                                 |   ✓   |   ✓   |   ✓   | Confirmation bottom-sheet; invalidates the session row                |
+| Log out ("Abmelden" / "Auf allen Geräten abmelden")     |   ✓   |   ✓   |   ✓   | Confirmation bottom-sheet; `DELETE /session` (this device) or `DELETE /sessions` (all devices). Nothing is deleted — account, PIN and uploads stay |
 | Rename own display name                                 |   ◯   |   ◯   |   ◯   | Not yet wired; PIN-protected change                                   |
 | Pick **data mode** (Saver / Original) in My Account     |   ✓   |   ✓   |   ✓   | Saver = compressed (default). Original = full files + data-usage warning. Applies to feed and diashow. Per-device, in `localStorage` |
 | Read the **Datenschutzhinweis** (privacy note)          |   ✓   |   ✓   |   ✓   | Free text set by Admin during setup; rendered preformatted in My Account; first-visit guide briefly points to it |
@@ -42,7 +42,7 @@ Status legend: **✓ shipped** · **◐ partial** · **◯ planned** · **✗ ou
 | 3-column grid feed with toggle                          |   ✓   |   ✓   |   ✓   | Video play badges, duration                                           |
 | Search & autocomplete (uploader + hashtag)              |   ✓   |   ✓   |   ✓   | Grid view; derived in-memory, no extra API calls                      |
 | Active filter chips (OR within type, AND across types)  |   ✓   |   ✓   |   ✓   | Multiple hashtags = OR; uploader + hashtag = AND                      |
-| Fullscreen lightbox with swipe                          |   ✓   |   ✓   |   ✓   | Swipe navigates the filtered set                                      |
+| Fullscreen lightbox with prev/next navigation           |   ✓   |   ✓   |   ✓   | Navigates the *filtered* set. On-screen prev/next controls, ← / → arrow keys, and swipe left/right on touch |
 | Like / unlike any post                                  |   ✓   |   ✓   |   ✓   | Single toggle; SSE `like-update`                                      |
 | Read comments on any post                               |   ✓   |   ✓   |   ✓   |                                                                       |
 | Add a comment                                           |   ✓   |   ✓   |   ✓   | Hashtags in comments also parsed                                      |
@@ -56,11 +56,11 @@ Status legend: **✓ shipped** · **◐ partial** · **◯ planned** · **✗ ou
 |                                                         |       |       |       |                                                                       |
 | **Moderation (Host)**                                   |       |       |       |                                                                       |
 | List all event users                                    |       |   ✓   |   ✓   | Includes upload count, total bytes                                    |
-| Ban / unban a user                                      |       |   ✓   |   ✓   | Modal asks: hide their existing uploads, or keep visible?             |
+| Ban / unban a user                                      |       |   ✓   |   ✓   | Ban **always hides** their existing uploads — no opt-out, the modal is a plain confirm. Unban restores them. Both invalidate and rebuild a released keepsake. See USER_JOURNEYS §9/§10 |
 | Delete any upload                                       |       |   ✓   |   ✓   |                                                                       |
 | Delete any comment                                      |       |   ✓   |   ✓   |                                                                       |
 | Promote guest to Host                                   |       |   ✓   |   ✓   |                                                                       |
-| Demote Host to guest                                    |       |   ✓   |   ✓   | Hosts may demote other Hosts. Cannot demote self. Admins cannot be demoted by hosts. |
+| Demote Host to guest                                    |       |       |   ✓   | **Admin only** — the backend 403s a Host demoting a peer Host (F1: it would launder past the ban / PIN-reset peer guards). Nobody may change their own role; Admins cannot be demoted at all. The button is hidden for non-admin Hosts |
 | Reset a guest's PIN (Host) / any non-admin PIN (Admin)  |       |   ✓   |   ✓   | New PIN shown once in modal; Host shows/shares it with the guest      |
 | Lock new uploads ("Event schließen")                    |       |   ✓   |   ✓   | Likes + comments + browsing remain open                               |
 | Unlock new uploads                                      |       |   ✓   |   ✓   |                                                                       |
@@ -78,18 +78,18 @@ Status legend: **✓ shipped** · **◐ partial** · **◯ planned** · **✗ ou
 | Edit compression-worker concurrency                     |       |       |   ✓   |                                                                       |
 | Edit **Datenschutzhinweis** (privacy note, free text)   |       |       |   ✓   | Plain text, whitespace + newlines preserved, no HTML. SSE `event-updated` broadcasts edits live. |
 | Inspect export job list & progress                      |       |       |   ✓   |                                                                       |
-| Low-disk alert (< 10 GB free)                           |       |       |   ◯   | Planned                                                               |
+| Low-disk alert                                          |       |   ✓   |   ✓   | Red banner at the top of the Host dashboard when `disk_low`, naming free space and the keepsake's required bytes. Fails closed to "not low" on an unreadable mount |
 | Event banner / cover image                              |       |       |   ◯   | DB column exists, no UI                                               |
 |                                                         |       |       |       |                                                                       |
-| **Quota visibility (Guest-facing)**                     |       |       |       |                                                                       |
-| Show current per-user quota estimate                    |   ✓   |   ✓   |   ✓   | "Du hast X MB von Y MB genutzt." in My Account and on the upload screen. Computed from the live formula. Hidden when quota enforcement is toggled off |
+| **Quota visibility (staff-only)**                       |       |       |       |                                                                       |
+| Show current per-user quota estimate                    |       |   ✓   |   ✓   | **Deliberately not shown to guests** — the storage widget in My Account and on the upload screen is gated on host/admin. Computed from the live formula; also hidden when quota enforcement is toggled off |
 |                                                         |       |       |       |                                                                       |
 | **Export**                                              |       |       |       |                                                                       |
 | Wait at locked export page until released               |   ✓   |   ✓   |   ✓   | Friendly "not yet available" copy                                     |
 | Download `Gallery.zip` (full-quality originals)         |   ✓   |   ✓   |   ✓   | Streamed via `async-zip`; `Photos/` + `Videos/` folders               |
 | Download `Memories.zip` (offline HTML viewer)           |   ✓   |   ✓   |   ✓   | Self-contained SvelteKit-static app + `data.json` + `media/`          |
 | HTML-export in-app guide modal before download          |   ✓   |   ✓   |   ✓   | Explains: unzip first, open `index.html`                              |
-| Per-IP export download rate limit (3 / day)             |   ✓   |   ✓   |   ✓   |                                                                       |
+| Per-**user** export download rate limit (3 / day)       |   ✓   |   ✓   |   ✓   | Keyed on the user id, not the IP — a venue behind one NAT would otherwise share a 3/day budget across every guest |
 |                                                         |       |       |       |                                                                       |
 | **Banned guest** (subset)                               |       |       |       |                                                                       |
 | Cannot upload, like, or comment                         |   ✗   |       |       | Returns HTTP 403                                                      |
@@ -143,7 +143,7 @@ email, no password, no account portal.
   PIN-recovery form for that account ("Already taken — sign in instead, or pick another
   name"). The join page also surfaces an explicit **"Ich habe bereits einen Account"**
   link routing to `/recover` for users who already know they want to sign in.
-- **PIN reset by Host / Admin.** Planned. If a guest loses their PIN and `localStorage` is
+- **PIN reset by Host / Admin.** If a guest loses their PIN and `localStorage` is
   gone everywhere, a Host (for guests) or Admin (for hosts and guests) can hit a
   **PIN zurücksetzen** action in the user list. A fresh PIN is generated server-side, its
   bcrypt stored, and the plaintext is shown **once** in a modal to the requesting
@@ -153,9 +153,13 @@ email, no password, no account portal.
   Admins can.
 - **Roles.** `guest` (default), `host`, `admin`. The Admin role is seeded from the
   `ADMIN_PASSWORD_HASH` env var; admins log in at `/admin/login` with a password (separate
-  JWT, 1-day expiry, in `sessionStorage`). Hosts are guests promoted by an admin. **Hosts
-  may also demote other Hosts to guests** (planned) — but never themselves, to avoid
-  locking the event out of moderation. Admins can demote anyone except admins.
+  JWT, 1-day expiry, in `sessionStorage`). Hosts are guests promoted by a host or an admin.
+  **Only an Admin may change a Host's role** — a plain Host may promote guests but may not
+  demote a peer Host, because demoting one would then let them ban or PIN-reset (→
+  `/recover` takeover) that ex-peer, the guards keying off the target's *current* role (F1).
+  Nobody may change their own role, so the event can't be locked out of moderation, and
+  Admins are un-demotable and un-bannable. Enforced in `handlers::host::set_role`; the
+  button is hidden for non-admin Hosts.
 
 ### 2.2 Posting pipeline
 
@@ -188,8 +192,9 @@ The upload pipeline is built for flaky mobile networks:
   the loaded uploads, so typing never hits the server.
 - **Filter chips** — multiple hashtags combine with OR; multiple uploaders combine with OR;
   hashtag + uploader combine with AND. Matches the redesign concept exactly.
-- **Lightbox** — fullscreen view, swipe navigates the *filtered* set, with embedded
-  like/comment UI.
+- **Lightbox** — fullscreen view with embedded like/comment UI. Navigation across the
+  *filtered* set is available three ways: on-screen prev/next controls, the ← / → arrow
+  keys, and a left/right swipe on touch.
 - **Real-time** — SSE delivers `new-upload`, `upload-processed`, `like-update`,
   `new-comment`, `upload-deleted`, `event-closed`/`event-opened`, `export-progress`,
   `export-available`. Client pauses SSE on `visibilitychange: hidden` and reopens on visible.
@@ -197,15 +202,18 @@ The upload pipeline is built for flaky mobile networks:
 ### 2.4 Host / Admin tooling
 
 - **Host dashboard** — three collapsible sections: Stats, Event-Einstellungen,
-  Nutzerverwaltung. Ban modal asks explicitly whether to hide the user's existing uploads
-  from the public feed. Promote/demote, lock/unlock, release-gallery are one-tap.
-- **Admin dashboard** — same dashboard plus three more inner tabs (Stats, Config, Export,
+  Nutzerverwaltung. The ban modal is a plain confirm: banning **always hides** the user's
+  existing uploads, there is no keep-visible option. Promote/demote/unban/release are each
+  behind a confirmation sheet that spells out the consequence, not one-tap.
+- **Admin dashboard** — same user list plus four inner tabs (Stats, Config, Export,
   Nutzer). Config form covers per-file limits, rate limits, quota tolerance, estimated
-  guest count, and compression concurrency — all stored in the `config` table and read on
+  guest count and the colour theme — all stored in the `config` table and read on
   each request, so changes take effect without a restart. Disk widget pulls from the
-  `sysinfo` crate live.
+  `sysinfo` crate live. The Export tab mirrors the host dashboard's release controls:
+  the release button is disabled once released, the live keepsake status (progress /
+  ready / failure reason) is shown, and a rebuild is offered as the recovery path.
 
-### 2.5 Data mode (planned)
+### 2.5 Data mode
 
 Each device picks a **data mode** in My Account; the setting lives in `localStorage` so a
 guest can be on Saver on their phone and Original on their laptop.
@@ -218,7 +226,7 @@ guest can be on Saver on their phone and Original on their laptop.
 Applies uniformly to the live app's feed/lightbox **and** the diashow. The viewer (offline
 HTML export) is unaffected — it's already a snapshot of pre-bundled media variants.
 
-### 2.6 Rate limits and quotas — toggleable (planned)
+### 2.6 Rate limits and quotas — toggleable
 
 The Admin Config tab gains explicit on/off toggles in addition to the numeric inputs:
 
@@ -230,13 +238,14 @@ The Admin Config tab gains explicit on/off toggles in addition to the numeric in
 - **Per-area quota switch.** Storage-bytes quota and upload-count quota can be disabled
   independently.
 
-When a feature is toggled off, the relevant UI in the guest-facing app should adapt: e.g.
-the "Du hast X von Y MB genutzt" widget hides itself when storage quota is disabled. The
-quota estimate is computed from the same formula the server uses
-(`(free_disk × tolerance) / max(active_uploaders, 1)`) — surfaced in My Account *and* on
-the upload preview screen so guests know before they pick files.
+When a feature is toggled off, the dependent UI adapts: the "Speicher: X / Y" widget hides
+itself when the storage quota is disabled. The quota estimate is computed from the same
+formula the server uses (`(free_disk × tolerance) / max(active_uploaders, 1)`) and is
+surfaced in My Account *and* on the upload preview screen — but **only to hosts and
+admins**. Guests never see it: the number moves as other people upload and as disk frees
+up, which reads as a broken or unfair limit to someone who can't see why.
 
-### 2.7 Privacy note (Datenschutzhinweis, planned)
+### 2.7 Privacy note (Datenschutzhinweis)
 
 Admin sets a free-text **Datenschutzhinweis** during instance setup (Admin Dashboard →
 Config). It's stored as a single config key (plain text, whitespace and newlines

docs/USER_JOURNEYS.md

@@ -19,8 +19,9 @@ can do what" overview, see [FEATURES.md](FEATURES.md). For manual QA, see
    this PIN is the only way to sign in on another device. PIN is also written to
    `localStorage`.
 6. Guest taps **Weiter zur Galerie** → lands in the feed (`/feed`).
-7. The **first-visit onboarding overlay** appears: dismissible steps (welcome, upload,
-   hashtags, PIN, and a brief pointer to the **Datenschutzhinweis** in My Account).
+7. The **first-visit onboarding overlay** appears: six dismissible steps (welcome, upload,
+   hashtags, long-press for more actions, hell/dunkel design pick, and PIN — the PIN step
+   also points at the **Datenschutzhinweis** in My Account when the admin has set one).
    `localStorage('eventsnap_guide_seen') = 'true'` after dismiss.
 8. Guest sees the bottom nav: **🏠 Feed · [📷+ FAB] · 👤 Account**.
 
@@ -41,7 +42,7 @@ can do what" overview, see [FEATURES.md](FEATURES.md). For manual QA, see
 5. Wrong PIN: up to 3 attempts. After the third, the account is locked for 15 minutes
    (`pin_locked_until` is set; further attempts return HTTP 429 with a localized message).
 
-## 4. PIN forgotten — Host or Admin resets it (planned)
+## 4. PIN forgotten — Host or Admin resets it
 
 The PIN is visible in **My Account** as long as `localStorage` is intact on at least one
 of the user's devices. If lost everywhere, the user asks a Host (or Admin) for a reset.
@@ -126,14 +127,23 @@ the Host can clean up later).
 3. **Event settings** — toggle to lock new uploads (likes / comments / browsing stay open;
    broadcasts `event-closed` SSE so all clients show a "uploads are locked" banner).
 4. **Galerie freigeben** — releases the export. Enqueues two export jobs (ZIP + HTML
-   viewer). Progress is visible in the Admin dashboard's Export tab; SSE
-   `export-progress` keeps it live; `export-available` notifies all guests when ready.
+   viewer). Progress is visible on the Host dashboard *and* in the Admin dashboard's Export
+   tab; SSE `export-progress` keeps it live; `export-available` notifies all guests when
+   ready. Once released the button is disabled and reads *"Galerie bereits freigegeben"* —
+   a second release is a 409. If a keepsake half fails, both dashboards show the reason and
+   offer **Erneut versuchen** (rebuild); a ready keepsake can also be rebuilt behind a
+   confirm, during which guest downloads are briefly unavailable.
 5. **Nutzerverwaltung** — search users; per-user controls:
    - **Sperren** opens a confirmation modal. Banning **always hides** the user's existing
      uploads (a banned user's content is "gone" everywhere) — there is no opt-out. Submitting
      calls `POST /host/users/{id}/ban` (no body).
-   - **Entsperren** lifts the ban. Same authority boundary as ban (below): a plain Host may
-     only unban Guests; only an Admin may unban a Host.
+   - **Entsperren** lifts the ban — and does two further things the confirm sheet now spells
+     out, because both are visible to the whole party: it clears `uploads_hidden`, so **all**
+     of that user's previously hidden uploads reappear in the gallery, the diashow and the
+     export; and it invalidates and rebuilds a released keepsake (`invalidate_and_arm`), so
+     every guest's download is unavailable for as long as that takes. A ban does the mirror
+     image of both. Same authority boundary as ban: a plain Host may only unban Guests; only
+     an Admin may unban a Host.
    - **Host** promotes a guest to host (Hosts and Admins may do this).
    - **Degradieren** — demote a Host back to guest. **Only an Admin may change a Host's
      role.** A plain Host may *not* demote a peer Host: doing so would let them then ban or
@@ -181,14 +191,18 @@ the Host can clean up later).
    - **Stats**: live counts and disk-usage widget (via `sysinfo`).
    - **Config**: per-file limits (image MB / video MB), rate limits (upload / feed /
      export), quota tolerance, estimated guest count, compression-worker concurrency,
-     plus the **Datenschutzhinweis** free-text editor and **on/off toggles** for the rate
-     limiters and quotas (planned — see §16). Whitelist on the server side rejects
-     unknown keys. Values are read from the `config` table on each request — no restart
-     needed.
-   - **Export**: list of past export jobs with status badges (pending / running / done /
-     failed) and progress bars; refresh button re-polls.
-   - **Nutzer**: same user list as Host, with the additional Demote action and (planned)
-     PIN-reset on host rows.
+     plus the colour-theme picker, the **Datenschutzhinweis** free-text editor and
+     **on/off toggles** for the rate limiters and quotas (see §18). Whitelist on the
+     server side rejects unknown keys. Values are read from the `config` table on each
+     request — no restart needed.
+   - **Export**: the gallery-release control plus the list of export jobs with status
+     badges (pending / running / done / failed) and progress bars; refresh button
+     re-polls. The release control mirrors the Host dashboard exactly — disabled once
+     released (a second tap would only 409), live keepsake status with progress and the
+     failure reason, and a **Neu erstellen / Erneut versuchen** rebuild as the recovery
+     path for a failed or stale keepsake.
+   - **Nutzer**: same user list as Host, with the additional Demote action (admin-only,
+     see §9) and PIN-reset on host rows.
 
 ## 12. Releasing the export and downloading
 
@@ -224,11 +238,14 @@ the Host can clean up later).
      buttons. Tapping the HTML download first shows an in-app guide modal explaining:
      "Entpacke die ZIP, öffne `index.html`". Tapping **Herunterladen** triggers the
      browser download.
-7. Downloads are rate-limited per IP (default 3 / day).
+7. Downloads are rate-limited **per user** (default 3 / day), keyed on the user id in
+   `enforce_export_rate` — deliberately not per IP. At a venue where a hundred guests share
+   one NAT'd wifi, a per-IP budget would be exhausted by the third person to tap
+   **Herunterladen** and lock everyone else out of their own keepsake.
 
-## 13. Diashow (planned)
+## 13. Diashow
 
-See [CONCEPT_DIASHOW.md](CONCEPT_DIASHOW.md). Summary of the planned flow:
+See [CONCEPT_DIASHOW.md](CONCEPT_DIASHOW.md). Summary of the flow:
 
 1. User taps a **Diashow / Präsentation** action (feed header on tablet/desktop, Account
    on mobile).
@@ -240,7 +257,7 @@ See [CONCEPT_DIASHOW.md](CONCEPT_DIASHOW.md). Summary of the planned flow:
    immediately.
 6. Tap or Escape reveals an overlay (pause, dwell selector, exit).
 
-## 14. Picking a data mode (planned)
+## 14. Picking a data mode
 
 1. Guest opens **My Account** → scrolls to **Datennutzung**.
 2. Two options: **Datensparer (empfohlen)** and **Original**. Saver is the default.
@@ -252,14 +269,24 @@ See [CONCEPT_DIASHOW.md](CONCEPT_DIASHOW.md). Summary of the planned flow:
 5. The viewer (offline HTML export) is unaffected — it already ships with its own pre-
    bundled `_thumb` / `_full` variants.
 
-## 15. Leaving an event
+## 15. Signing out
 
-1. User opens **My Account** → taps **🚪 Event verlassen**.
-2. Bottom-sheet confirmation: "Event verlassen?" with **Abmelden** and **Bleiben**.
-3. Confirming calls `DELETE /api/v1/session` (invalidates the session row), clears the JWT
-   and PIN from `localStorage`, and redirects to the join page.
+The wording matters here: nothing is *left* and nothing is deleted. The account, the PIN
+and every uploaded photo survive — the user signs back in with name + PIN whenever they
+like. ("Event verlassen" read as leaving for good, which the confirmation sheet then
+contradicted by saying you can come back.)
 
-## 16. Reading the Datenschutzhinweis (planned)
+1. User opens **My Account** → Konto section. Two separate actions:
+   - **Abmelden** — this device only (`DELETE /api/v1/session`).
+   - **Auf allen Geräten abmelden** — every session of theirs (`DELETE /api/v1/sessions`),
+     for a lost or borrowed phone.
+2. Bottom-sheet confirmation naming which of the two it is, with **Abmelden** / **Abbrechen**.
+3. Confirming invalidates the session row(s), wipes the local JWT, PIN and the IndexedDB
+   upload queue (so the next guest on a shared phone doesn't inherit pending uploads), and
+   redirects to the join page. The session-delete call is best-effort: the token is gone
+   locally either way, so a network failure never traps the user on the page.
+
+## 16. Reading the Datenschutzhinweis
 
 1. User opens **My Account** → scrolls to **Datenschutzhinweis**.
 2. The note is rendered inside a preformatted block (`<pre>`-style: monospace, whitespace
@@ -270,11 +297,15 @@ See [CONCEPT_DIASHOW.md](CONCEPT_DIASHOW.md). Summary of the planned flow:
 4. Admin sets / edits the note in **Admin Dashboard → Config → Datenschutzhinweis**: a
    tall textarea with a save button. Saved to a single `config` key.
 
-## 17. Mobile-first gestures (planned)
+## 17. Mobile-first gestures (partly shipped)
 
 EventSnap's UI is mobile-first; gestures replace explicit buttons where they're more
 ergonomic. Buttons are always present as fallback for desktop and accessibility.
 
+Shipped today: long-press context sheets on posts and comments, and lightbox navigation
+(on-screen prev/next controls, ← / → arrow keys, and left/right swipe). The remaining rows
+below are still planned.
+
 | Gesture                                   | Action                                                |
 |-------------------------------------------|-------------------------------------------------------|
 | Long-press on a post (own)                | Bottom sheet → Löschen, Original anzeigen, Teilen     |
@@ -294,7 +325,7 @@ Inspiration: Instagram (double-tap heart, swipe stories), WhatsApp (long-press f
 context), Telegram (swipe-to-reply on messages — could inform comment threads if those
 land).
 
-## 18. Admin toggles a rate limit or quota off (planned)
+## 18. Admin toggles a rate limit or quota off
 
 1. Admin opens **Admin Dashboard → Config**.
 2. **Rate-Limits** section: a master switch and per-endpoint switches (upload / feed /
@@ -305,8 +336,9 @@ land).
    limiter entirely.
 5. **Quoten** section mirrors the pattern: master toggle plus per-area toggles (storage
    bytes / upload count).
-6. When the storage-quota toggle is off, the **"X von Y MB genutzt"** widget in the
-   guest's My Account and upload screen hides itself (no quota → no number to show).
+6. When the storage-quota toggle is off, the **"Speicher: X / Y"** widget in My Account and
+   on the upload screen hides itself (no quota → no number to show). That widget is
+   staff-only in any case — guests never see it.
 
 Suggested defaults at deploy time: all toggles **on**, sensible numeric limits.
 Toggling off is the explicit escape hatch for testing or trusted internal events.

e2e/Caddyfile.test

@@ -12,10 +12,17 @@
     # Mirror prod's security headers (minus HSTS, which is HTTPS-only).
     header {
         X-Content-Type-Options "nosniff"
-        X-Frame-Options "DENY"
         Referrer-Policy "strict-origin-when-cross-origin"
     }
 
+    # Mirror prod's export carve-out: the keepsake download targets a hidden
+    # same-origin iframe, and WebKit enforces XFO before Content-Disposition.
+    # Two disjoint matchers, not an override — see the comment in ../Caddyfile.
+    @framable path /api/v1/export/zip /api/v1/export/html
+    @not_framable not path /api/v1/export/zip /api/v1/export/html
+    header @framable X-Frame-Options "SAMEORIGIN"
+    header @not_framable X-Frame-Options "DENY"
+
     reverse_proxy /api/* app:3000
     reverse_proxy /media/* app:3000
     reverse_proxy /health app:3000

e2e/docker-compose.test.yml

@@ -42,11 +42,29 @@ services:
       EVENT_NAME: E2E Test Event
       APP_PORT: '3000'
       MEDIA_PATH: /media
+      # Exports MUST live outside MEDIA_PATH — see the note on the volume below and
+      # config.rs::validate. Omitting this left exports on the container's writable
+      # layer at the /exports default, so the test stack diverged from the prod layout
+      # it claims to mirror, and export-leak/export-video wrote real archives into
+      # ephemeral storage.
+      EXPORT_PATH: /exports
       SESSION_EXPIRY_DAYS: '30'
       EVENTSNAP_TEST_MODE: '1' # ENABLES /admin/__truncate — never set in prod
       RUST_LOG: eventsnap_backend=info,tower_http=warn
     volumes:
       - media_data:/media
+      # Separate volume, exactly as in production: a keepsake archive contains every
+      # photo in the event, so it is kept off the media tree.
+      - exports_data:/exports
+    # Mirror production's cap (docker-compose.yml). The test stack having NO memory limit is
+    # why an unbounded image decode was invisible here: a 99 MP upload that would OOM-kill the
+    # 1 GiB production container simply succeeded in CI. A test environment more generous than
+    # production cannot catch a resource bug — the same shape as WebKit being absent from CI
+    # and /health existing only in Caddyfile.test.
+    deploy:
+      resources:
+        limits:
+          memory: 1G
     expose:
       - '3000'
 
@@ -75,3 +93,4 @@ services:
 
 volumes:
   media_data:
+  exports_data:

e2e/fixtures/db.ts

@@ -37,6 +37,52 @@ export const db = {
     );
   },
 
+  /**
+   * Is this user's account currently PIN-locked?
+   *
+   * Distinguishes the two ways /recover can answer 429 — the per-(IP, name) throttle, which
+   * costs the attacker, and the account lock, which costs the VICTIM. Only the second one is
+   * weaponizable, so a test asserting "a single IP cannot lock a guest out" has to look at the
+   * row, not at the status code.
+   */
+  async isPinLocked(userId: string): Promise<boolean> {
+    return withClient(async (c) => {
+      const r = await c.query<{ locked: boolean }>(
+        `SELECT (pin_locked_until IS NOT NULL AND pin_locked_until > NOW()) AS locked
+           FROM "user" WHERE id = $1`,
+        [userId]
+      );
+      return r.rows[0]?.locked ?? false;
+    });
+  },
+
+  /**
+   * Preload the wrong-PIN streak, standing in for failures that arrived from other IPs.
+   *
+   * The account lock is deliberately out of reach of any single source, so a test that wants to
+   * exercise it has to simulate the distributed case rather than hammer from one address.
+   * `last_failed_pin_at` is set to now so the 15-minute decay does not immediately reset it.
+   */
+  async setFailedPinAttempts(userId: string, attempts: number) {
+    await withClient((c) =>
+      c.query(
+        `UPDATE "user" SET failed_pin_attempts = $2, last_failed_pin_at = NOW() WHERE id = $1`,
+        [userId, attempts]
+      )
+    );
+  },
+
+  /** Current wrong-PIN streak. Decays after 15 minutes — see User::increment_failed_pin. */
+  async failedPinAttempts(userId: string): Promise<number> {
+    return withClient(async (c) => {
+      const r = await c.query<{ failed_pin_attempts: number }>(
+        `SELECT failed_pin_attempts FROM "user" WHERE id = $1`,
+        [userId]
+      );
+      return r.rows[0]?.failed_pin_attempts ?? 0;
+    });
+  },
+
   async expireSession(userId: string) {
     await withClient((c) =>
       c.query(`UPDATE session SET expires_at = NOW() - interval '1 hour' WHERE user_id = $1`, [
@@ -54,6 +100,27 @@ export const db = {
     );
   },
 
+  async compressionStatus(uploadId: string): Promise<string | null> {
+    return withClient(async (c) => {
+      const r = await c.query<{ compression_status: string }>(
+        `SELECT compression_status FROM upload WHERE id = $1`,
+        [uploadId]
+      );
+      return r.rows[0]?.compression_status ?? null;
+    });
+  },
+
+  /** Which revision of the derivative pipeline produced this row's preview/display. */
+  async derivativesRev(uploadId: string): Promise<number | null> {
+    return withClient(async (c) => {
+      const r = await c.query<{ derivatives_rev: number }>(
+        `SELECT derivatives_rev FROM upload WHERE id = $1`,
+        [uploadId]
+      );
+      return r.rows[0]?.derivatives_rev ?? null;
+    });
+  },
+
   async countUploadsForUser(userId: string): Promise<number> {
     return withClient(async (c) => {
       const r = await c.query<{ count: string }>(
@@ -84,6 +151,20 @@ export const db = {
     });
   },
 
+  /**
+   * Overstate an upload's recorded size.
+   *
+   * The keepsake size estimate and the low-disk threshold are pure SQL over
+   * `original_size_bytes` — no file is read — so this is the lever for driving "the keepsake
+   * would not fit" without a genuinely full disk. The bytes on disk are unchanged; only the
+   * accounting the warning reads from moves.
+   */
+  async setUploadSizeBytes(uploadId: string, bytes: number) {
+    await withClient((c) =>
+      c.query(`UPDATE upload SET original_size_bytes = $2 WHERE id = $1`, [uploadId, bytes])
+    );
+  },
+
   async setExportReleased(slug: string, released: boolean) {
     await withClient((c) =>
       c.query(`UPDATE event SET export_released_at = $2 WHERE slug = $1`, [
@@ -121,7 +202,8 @@ export const db = {
   async fakeExportJob(
     eventSlug: string,
     type: 'zip' | 'html',
-    status: 'pending' | 'running' | 'done'
+    status: 'pending' | 'running' | 'done' | 'failed',
+    errorMessage: string | null = null
   ) {
     await withClient(async (c) => {
       const ev = await c.query<{ id: string; export_epoch: string }>(
@@ -130,11 +212,12 @@ export const db = {
       );
       if (ev.rows.length === 0) throw new Error(`No event with slug ${eventSlug}`);
       await c.query(
-        `INSERT INTO export_job (event_id, type, status, progress_pct, completed_at, epoch)
-         VALUES ($1, $2::export_type, $3::export_status, $4, $5, $6)
+        `INSERT INTO export_job (event_id, type, status, progress_pct, completed_at, epoch,
+                                 error_message)
+         VALUES ($1, $2::export_type, $3::export_status, $4, $5, $6, $7)
          ON CONFLICT (event_id, type) DO UPDATE
            SET status = EXCLUDED.status, progress_pct = EXCLUDED.progress_pct,
-               epoch = EXCLUDED.epoch`,
+               epoch = EXCLUDED.epoch, error_message = EXCLUDED.error_message`,
         [
           ev.rows[0].id,
           type,
@@ -142,6 +225,7 @@ export const db = {
           status === 'done' ? 100 : 0,
           status === 'done' ? new Date() : null,
           ev.rows[0].export_epoch,
+          errorMessage,
         ]
       );
     });

e2e/fixtures/media/not-an-image.jpg

@@ -0,0 +1,4 @@
+This is plain text, not an image at all.
+This is plain text, not an image at all.
+This is plain text, not an image at all.
+This is plain text, not an image at all.

e2e/helpers/webkit.ts

@@ -0,0 +1,29 @@
+import { test } from '@playwright/test';
+
+/**
+ * Skip a test that depends on persisting a Blob/File in IndexedDB when running on
+ * Playwright's WebKit.
+ *
+ * The client upload queue (`frontend/src/lib/upload-queue.ts`) stores the file itself in
+ * IndexedDB so a backgrounded or reloaded phone can resume the upload. Playwright's Linux
+ * WebKit build cannot store Blobs in IndexedDB at all — `put()` fails with
+ * "UnknownError: Error preparing Blob/File data to be stored in object store". Verified to
+ * be the harness, not the app: a Blob constructed in-page with `new Blob([bytes])` fails
+ * exactly the same way, while Chromium stores both that and a `setInputFiles` File fine.
+ * Real iOS Safari supports Blobs in IndexedDB, so this is NOT evidence of a bug on the
+ * platform these tests exist to protect.
+ *
+ * Any test that drives the composer (FAB → UploadSheet → /upload → submit) hits this,
+ * because `handleSubmit` awaits `addToQueue`, which throws before it can navigate.
+ *
+ * This is deliberately narrow. WebKit still runs every API-driven upload test, the whole of
+ * 01-auth, 03-feed and 06-export — including the keepsake download, which only WebKit can
+ * meaningfully verify. If Playwright's WebKit ever gains IndexedDB Blob support, delete this
+ * helper and the four call sites.
+ */
+export function skipIfNoIdbBlobs(browserName: string) {
+  test.skip(
+    browserName === 'webkit',
+    "Playwright's Linux WebKit cannot store Blobs in IndexedDB (harness limitation, not an iOS one) — the client upload queue can't be exercised there"
+  );
+}

e2e/loadtest/.gitignore

@@ -0,0 +1,4 @@
+results/
+__pycache__/
+*.pyc
+*.log

e2e/loadtest/README.md

@@ -0,0 +1,112 @@
+# EventSnap load / stress test
+
+Simulates a real event: **~100 guests** joining and uploading **~1000 images** in
+bursts (10–20 at a time) spread across a compressed time window, a pool of
+**viewers** holding live SSE connections, and **one real browser** on `/diashow`
+acting as the showcase display.
+
+## Goal (this harness is tuned for it)
+
+**Validate the shipping config.** We run at the real production defaults —
+compression concurrency (`COMPRESSION_WORKER_CONCURRENCY`, default **2**), DB
+pool (default **10**), quotas **on** — and answer: _does the app survive the
+event, and how far behind real-time does the diashow fall?_
+
+The headline metric is **pipeline latency**: time from an upload succeeding to
+its preview being ready (`upload-processed` SSE event) — i.e. _how long until the
+photo appears on the diashow_. A backlog that builds is fine; a backlog that
+**never drains** is a fail for a live event.
+
+## Methodology: what we change vs. shipping
+
+We **only disable rate limits**. They're per-IP / per-user anti-abuse guards; a
+synthetic test from one IP trips them in a way real guests (distinct IPs, phones)
+never would — leaving them on would measure the limiter, not the pipeline.
+Everything else (compression concurrency, DB pool, quotas) stays at the real
+default so the result is honest.
+
+> **Standalone finding to remember:** the shipping `upload_rate_per_hour` default
+> is **10**. A real guest uploading a burst of 10–20 photos would be throttled by
+> the shipping config too. That's a genuine event-day issue worth surfacing
+> separately from this pipeline test.
+
+## Prereqs
+
+- The isolated test stack up: `cd e2e && npm run stack:up` (Caddy on `:3101`,
+  `EVENTSNAP_TEST_MODE=1`, `/admin/__truncate` live).
+- Node 24+ (global `fetch`/`FormData`/`Blob`), Python 3 + Pillow, Docker CLI
+  access (used for `docker stats` + `docker exec psql` ground-truth sampling).
+- `@playwright/test` (already an e2e dep) for the diashow watcher.
+
+## 1. Generate the image pool (once)
+
+Realistic phone-sized JPEGs (~2–4 MB, 12 MP, high entropy). The driver reuses
+this pool at random across all 1000 uploads — real load is byte size + decode
+cost, not file uniqueness.
+
+```bash
+python3 e2e/loadtest/gen-images.py 40      # → /tmp/eventsnap-loadtest/photos
+```
+
+~40 images ≈ 120 MB pool; projects to **~3–4 GB** of originals for 1000 uploads
+(previews/thumbnails add more). The generator prints the projection; check disk.
+
+## 2. Smoke run first (~1 min)
+
+Proves the wiring — join, upload, SSE correlation, drain, metrics — before the
+real thing:
+
+```bash
+LT_GUESTS=5 LT_IMAGES=50 LT_WINDOW_SEC=60 node e2e/loadtest/driver.mjs
+```
+
+## 3. Full run (~15 min + drain)
+
+Two terminals. Start the showcase display first, then the driver:
+
+```bash
+# terminal A — the showcase device
+node e2e/loadtest/diashow-watch.mjs
+
+# terminal B — 100 guests / 1000 images / 15-min window (defaults)
+node e2e/loadtest/driver.mjs
+```
+
+The driver truncates event data first (`LT_TRUNCATE=0` to keep), disables rate
+limits, joins guests, opens SSE, runs the burst schedule, then **waits for the
+compression backlog to drain** before reporting.
+
+## Output
+
+- Console: live progress every 10 s, then a RESULTS block with pass/fail flags.
+- `e2e/loadtest/results/run-<timestamp>.json`: full metrics — upload latency
+  percentiles, pipeline latency percentiles, drain time, per-status counts, SSE
+  reconnect/resync counts, and a `docker stats` + DB-connection time series.
+- `e2e/loadtest/results/diashow/`: periodic screenshots of the live display.
+
+## What the flags mean
+
+| Flag                      | Meaning                                                                      |
+| ------------------------- | ---------------------------------------------------------------------------- |
+| `✗ 5xx`                   | server errored under load — hard fail                                        |
+| `✗ 507`                   | quota rejected uploads — disk/quota misconfig for the event size             |
+| `✗ backlog did not drain` | compression can't keep up even after uploads stop — diashow never catches up |
+| `⚠ pipeline p95 > 60s`    | photos take >1 min to appear on the diashow at peak                          |
+| `⚠ SSE resyncs`           | live consumers lagged the broadcast channel                                  |
+
+## Knobs
+
+All via env (see header of `driver.mjs`): `LT_GUESTS`, `LT_IMAGES`,
+`LT_WINDOW_SEC`, `LT_BURST_MIN/MAX`, `LT_BURST_CONC`, `LT_VIEWERS`,
+`LT_TRUNCATE`, `LT_DRAIN_TIMEOUT_SEC`, `LT_KEEP_RATELIMITS`, `LT_BASE`,
+`LT_APP_CONTAINER`, `LT_DB_CONTAINER`.
+
+To later answer _"what config should I deploy?"_, re-run with a rebuilt stack
+that sets `COMPRESSION_WORKER_CONCURRENCY` higher (boot-time env var in
+`docker-compose.test.yml`) and compare the pipeline-latency / drain numbers.
+
+## Teardown
+
+```bash
+cd e2e && npm run stack:down     # wipes volumes (media + db)
+```