fix: gate uploads on keepsake headroom, and close five unattended-event gaps

The box is 2 vCPU / 4 GB / 40 GB, not the 4 vCPU / 8 GB / 80 GB that the audit,
the committed comments and README's sizing section all assumed. That correction
is what the first change is about; the rest are the remaining pre-event items.

THE ARCHIVE COULD BECOME UNBUILDABLE WHILE UPLOADS KEPT SUCCEEDING

`required_free_bytes` is `media × 1.1 × 2` — the ZIP and the HTML viewer are each
gallery-sized — and the export preflight also wants DISK_RESERVE_BYTES on top. The
upload gate, though, only refused below a FLAT 10 GB reserve. On 40 GB that let
uploads run to ~25 GB of media while a release needed `2.2 × 25 + 10` = 65 GB free.
Every upload in that band succeeded and the keepsake could then never be built: the
product's entire promise, failing silently at the end of the night with nobody there.

The gate now enforces the invariant that actually matters — never accept an upload
that would make the keepsake unbuildable — sharing `required_free_bytes` with the
preflight so the two cannot drift into disagreeing about the same question. Uploads
stop at ~8 GB of media on this disk, with a German message naming the cause.
Refusing the 1001st photo beats losing all 1000.

`media_total.rs` backs it: SUM(user.total_upload_bytes) over ~100 rows, cached 5s,
rather than `estimate_export_bytes`'s join across every upload. It counts hidden and
banned users' bytes, which the export excludes — skew in the SAFE direction, so the
gate closes marginally early rather than late. Fails open on a query error.

A test pins the gate against the preflight across the whole gallery-size range, and
a second asserts the per-user floor alone would over-commit the volume — i.e. that
the global gate is what must bind.

THE WATCHDOG ABORTED HEALTHY UPLOADS EVERY TIME A PHONE WAS POCKETED

`Date.now()` advances while a backgrounded phone is frozen but `setInterval` does
not, so the first tick after a screen lock read the whole sleep as silence and
aborted — re-sending a video from byte zero and burning one of five PERMANENT
auto-attempts. The interval is now its own suspension detector: a tick that arrives
125s late for a 5s schedule credits that window back, because a period the watchdog
could not observe is not evidence of silence.

Chosen over a `visibilitychange` listener, which only covers causes that fire that
event — a throttled-but-visible tab, a closed lid and an occluded window all freeze
timers without one — and which would have needed module state, an SSR guard and a
teardown for strictly less coverage. `performance.now()` was rejected because Safari
pauses it across system sleep on some paths and Chrome does not.

The credit buys one fresh window, not immunity: a socket iOS reaped while
backgrounded still aborts ~90s after resume rather than hanging for `xhr.timeout`
(5-60 min) with the queue's `processing` latch held.

Two latent leaks found while in there: `xhr.abort()` on a request already in
readyState DONE emits no `abort` event, so `settle()` never ran and the interval
re-aborted every 5s forever while `activeUploads` kept a stale entry (the ✕ button
silently stopped working); and a synchronous throw from `xhr.send` — a blob whose
backing store the OS purged — leaked the same way. Both closed.

OKLCH MADE THE DELETE BUTTON INVISIBLE ON SAMSUNG'S DEFAULT BROWSER

red/amber/green were never in the @theme block and fell through to Tailwind v4's
`oklch()` defaults, which Safari <15.4, Chrome <111 and Samsung Internet <22 cannot
parse: `var(--color-red-600)` is then invalid at computed-value time, `background-color`
falls back to transparent, and `.btn-danger` renders white text on nothing. Pinned to
Tailwind's own defaults gamut-mapped to sRGB by Lightning CSS — the converter already
in this pipeline — so modern browsers render exactly what they render today. Verified
against seven hex fallbacks it had already emitted for the /alpha forms. rose and teal
(avatar chips) had the same leak. The app CSS goes from 40 oklch declarations to 0.

Also fixes `--color-purple-950`, which was simply missing: `dark:bg-purple-950/50` on
the host dashboard was rendering default violet on EVERY browser, off-brand.

The keepsake viewer only picks this up on a rebuild, so its committed artefact is
rebuilt here too — still single-file, still zero external references.

A BRICKED BOOT LOOKED LIKE A SPINNER FOREVER

With `ssr = false` the page is empty until the bundle mounts, so a chunk 404 after a
redeploy or a dead uplink left the guest on the boot spinner with no message, no
reload control, and in a standalone PWA no URL bar. A 15s timeout in the existing
nonce'd IIFE (no CSP change) swaps in German copy and a reload button. Deliberately a
timeout rather than feature detection: a SyntaxError in the bundle is invisible to any
capability check. Plus a <noscript>, since there was nothing at all to see without JS.

EVERY 4xx WAS INVISIBLE AT ANY LOG LEVEL

tower_http counts 4xx as a success, so it logs at DEBUG while production runs at info.
If guests spend the evening hitting 429s or 413s, the post-event logs said nothing.
Now one WARN per client error; 5xx excluded because Internal already logs its source
chain and the pool-exhaustion 503 logs at construction.

A DEAD FRONTEND SERVED A BLANK 502

`handle_errors 5xx` with an inline German page (the caddy service mounts only the
Caddyfile, so there is no volume to ship a static file through). Verified empirically
against this config, not from documentation: an upstream 404 through `reverse_proxy`
still arrives as untouched `application/json`, and only a dial failure renders the
page. That mattered — the keepsake download navigates a hidden iframe and DEPENDS on a
real 404/429 arriving, and swallowing those would have been worse than the blank 502.

CONFIG CORRECTIONS FOR THE REAL HARDWARE

DATABASE_MAX_CONNECTIONS 30 → 15: sized to 2 vCPU rather than to the guest count.
Since migration 024 a feed page costs well under a millisecond, so connections are no
longer spent waiting, and 30 backends crowd the db container's 1 GB on a 4 GB host.
COMPRESSION_WORKER_CONCURRENCY stays at 2 — the merged heavy-image permit already
serialises anything over 150 MiB, so the "two 48 MP photos" worst case that number was
sized against is unreachable; dropping to 1 would halve light-path throughput and push
more feed tiles onto full-size originals. README's sizing section rewritten for the
actual disk.

Verified: 149/149 backend tests against a live Postgres, clippy clean, 57/57 vitest,
svelte-check 0 errors, eslint clean, vite build, export-viewer rebuild, caddy validate,
compose YAML parse.

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

.env.example

@@ -36,13 +36,20 @@ DATABASE_URL=postgres://eventsnap:CHANGE_ME_use_a_strong_password@db:5432/events
 POSTGRES_USER=eventsnap
 POSTGRES_PASSWORD=CHANGE_ME_use_a_strong_password
 POSTGRES_DB=eventsnap
-# Connection pool size. Default 10. For a busy event (~100 guests polling the feed
+# Connection pool size. The code default is 10 (backend/src/db.rs) — set it explicitly,
-# + SSE + uploads at once) raise to ~30 so requests don't queue on a pool permit.
+# because a `.env` written by hand from this file's secrets is otherwise silently on 10.
-# PAIRED WITH THE DB CONTAINER'S MEMORY LIMIT: 30 backends plus Postgres 16's default
+#
-# shared_buffers is already snug in the 1G that docker-compose.yml allots the `db`
+# SIZE IT TO THE CORES, NOT TO THE GUESTS. The earlier advice here was ~30, reasoned from
-# service. If you raise this, raise `db.deploy.resources.limits.memory` with it — an
+# "~100 guests polling the feed at once" back when a feed page cost ~449 ms and connections
-# OOM in Postgres doesn't degrade one feature, it takes the whole event down.
+# were spent waiting. Migration 024 replaced the feed view's GROUP BY with scalar subqueries
-DATABASE_MAX_CONNECTIONS=30
+# 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. `info` is the right production default: at `debug` the tower-http trace
 # layer writes a line per request AND per response, which on a busy event is a large
@@ -124,12 +131,23 @@ EXPORT_PATH=/exports
 # 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 2, two 48 MP photos ≈ 800 MB against the 1G app limit — ~25% margin.
+#   * concurrency 4 would put two giants at ~1.5 GB against the 1G app limit — OOM.
-#   * concurrency 4, the same pair ≈ 1.5 GB — 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

Caddyfile

@@ -75,4 +75,59 @@
 
 	# 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}
+	}
 }

README.md

@@ -280,35 +280,49 @@ so a host takedown or a ban actually revokes access to the bytes.
 degrade one subsystem — Postgres stops being able to write and the whole event goes
 down.
 
-Uploads are self-limiting. `per_user_limit = free_disk × quota_tolerance ÷
+**`Gallery.zip` and `Memories.zip` are each roughly a second copy of every original.**
-active_uploaders` is recomputed against live free space on every upload, so guests
+Both write their media `Compression::Stored`, and `Memories.zip` streams the untouched
-converge on a fixed point at `tolerance / (1 + tolerance)` of the free space you
+original for every video and for every image at or under 5 MB. So a release wants room
-started with — **43%** at the default 0.75. On an 80 GB box with ~70 GB free after
+for **two more copies of the gallery** on top of the gallery itself — which is what
-the OS and images, media settles at ~30 GB and stops.
+`required_free_bytes` encodes as `media × 1.1 × 2`.
 
-**The keepsake is what the 80 GB baseline does not cover.** `Gallery.zip` and
+The per-user quota does **not** bound this. It is a fairness mechanism that divides
-`Memories.zip` are built concurrently and each is roughly a second copy of every
+free space between guests, and since it carries a floor (`MIN_QUOTA_LIMIT_BYTES`, so a
-original: both write their media `Compression::Stored`, and `Memories.zip` streams the
+guest's allowance stops shrinking as the party fills up) the aggregate ceiling it used
-untouched original for every video and for every image at or under 5 MB. So a release
+to imply is gone. What bounds the disk is the **global gate in the upload handler**,
-wants room for **two more copies of the gallery** on top of the gallery itself.
+which refuses any upload that would leave too little room to build the keepsake:
 
-| Stage | Used | Free (80 GB box) |
+```
-|---|---|---|
+free_after_upload  <  media_after × 1.1 × 2  +  DISK_RESERVE_BYTES   →   refused
-| Fresh box (OS + images) | ~10 GB | ~70 GB |
+```
-| Guests reach the quota fixed point | ~40 GB | ~40 GB |
-| Host releases → both archives | ~100 GB | **ENOSPC** |
 
-Two ways to size for it:
+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.
 
-This is no longer silent. The export refuses up front with the two numbers rather than
+None of this is silent. The upload gate refuses with a German message naming the cause,
-hitting ENOSPC halfway through a multi-GB write, a rebuild reclaims the superseded
+the export preflight refuses up front with both numbers rather than hitting ENOSPC
-generation before it starts (so peak is one generation, not two), and the host
+halfway through a multi-GB write, a rebuild only reclaims the superseded generation
-dashboard warns as soon as the keepsake would not fit — which is the only point at
+**after** the new one lands (so a failed rebuild can never leave you with no archive at
-which anyone can still do something about it.
+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.
 
 ---
 

backend/src/error.rs

@@ -78,6 +78,26 @@ impl IntoResponse for AppError {
         };
         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.
+        if status.is_client_error() {
+            tracing::warn!(status = status.as_u16(), code, %message, "request rejected");
+        }
+
         let mut body = serde_json::json!({
             "error": code,
             "message": message,
@@ -161,6 +181,43 @@ mod tests {
         }
     }
 
+    /// 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]

backend/src/handlers/test_admin.rs

@@ -111,6 +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/handlers/upload.rs

@@ -438,43 +438,62 @@ pub async fn upload(
     let quota_on = config::get_bool(&state.config_cache, "quota_enabled", true).await;
     let storage_quota_on =
         config::get_bool(&state.config_cache, "storage_quota_enabled", true).await;
-    // When quota is enforced, this holds the byte ceiling so the increment UPDATE below can
+    // GLOBAL DISK GATE, checked before the per-user ceiling and independent of every quota
-    // enforce it atomically (`WHERE total + size <= limit`). Without that guard, two
-    // concurrent uploads from the same user (e.g. phone + laptop) both pass this stale
-    // pre-check and both increment, blowing past the quota. The pre-check stays as a
-    // fast path that avoids the disk write when the user is already clearly over.
-    // GLOBAL RESERVE, checked before the per-user ceiling and independent of every quota
     // toggle. The per-user quota is a fairness mechanism, not a disk guarantee — and since it
-    // now carries a floor (MIN_QUOTA_LIMIT_BYTES) so a guest's allowance stops shrinking as
+    // carries a floor (MIN_QUOTA_LIMIT_BYTES) so a guest's allowance stops shrinking as the
-    // the party fills up, the aggregate ceiling it used to imply is gone entirely. Something
+    // party fills up, the aggregate ceiling it used to imply is gone entirely. Something has to
-    // has to own "do not fill the volume", because `postgres_data`, `media_data` and
+    // own "do not fill the volume", because `postgres_data`, `media_data` and `exports_data`
-    // `exports_data` share one filesystem: the end state is not a degraded feature, it is
+    // share one filesystem: the end state is not a degraded feature, it is Postgres unable to
-    // Postgres unable to write WAL and the whole event down with nobody watching.
+    // write WAL and the whole event down with nobody watching.
     //
-    // Deliberately NOT gated behind `quota_enabled`. That switch exists so an operator can
+    // WHAT IS RESERVED IS NOT A CONSTANT. A flat reserve answers "can Postgres still write",
-    // stop rationing space between guests; it was never meant to authorise running the disk
+    // which is necessary and not sufficient: the keepsake needs room for BOTH halves at once —
-    // to zero, and an operator flipping it at 23:00 to unblock a guest should not silently
+    // `required_free_bytes` is `media × 1.1 × 2`, since the ZIP and the HTML viewer are each
-    // disarm the last thing standing between the party and a dead database.
+    // gallery-sized. On the 40 GB box this runs on, a flat 10 GB reserve let uploads continue to
+    // roughly 25 GB of media while the release needed `2.2 × 25 + 10` = 65 GB free. Every upload
+    // in that band succeeded and then the archive could never be built — the product's entire
+    // promise, failing silently at the end of the night with nobody there to notice.
+    //
+    // So the gate enforces the invariant that actually matters: never accept an upload that
+    // would make the keepsake unbuildable. It shares `required_free_bytes` with the export
+    // preflight so the two cannot drift into disagreeing about the same question.
+    //
+    // Deliberately NOT gated behind `quota_enabled`. That switch exists so an operator can stop
+    // rationing space between guests; it was never meant to authorise running the disk to zero,
+    // and an operator flipping it at 23:00 to unblock a guest should not silently disarm the
+    // last thing standing between the party and a dead database.
     if let Some(free) = crate::services::disk::free_bytes(&state.config.media_path) {
-        let remaining = (free as i64).saturating_sub(size);
+        let media_after = state.media_total.get(&state.pool).await.saturating_add(size);
-        if remaining < DISK_RESERVE_BYTES {
+        let keepsake_needs =
+            crate::services::export::required_free_bytes(media_after.max(0) as u64, 2) as i64;
+        let free_after = (free as i64).saturating_sub(size);
+        let required = keepsake_needs.saturating_add(DISK_RESERVE_BYTES);
+        if free_after < required {
             tracing::error!(
                 free_bytes = free,
                 upload_size = size,
+                media_after,
+                keepsake_needs,
                 reserve = DISK_RESERVE_BYTES,
-                "refusing upload: it would take the media volume below the reserve"
+                "refusing upload: it would leave too little room to build the keepsake"
             );
             return Err(AppError::QuotaExceeded(
-                "Der Speicher des Events ist voll. Bitte sag einem Host Bescheid — neue \
+                "Der Speicher des Events ist fast voll — damit die Galerie am Ende noch als \
-                 Uploads sind vorübergehend nicht möglich."
+                 Download erstellt werden kann, sind neue Uploads jetzt gesperrt. Bitte sag \
+                 einem Host Bescheid."
                     .into(),
             ));
         }
     }
     // Failing OPEN when the disk can't be read is deliberate and matches the per-user quota
-    // above: refusing every upload because a `statfs` failed would be a worse outage than the
+    // below: refusing every upload because a `statfs` failed would be a worse outage than the
     // one being guarded against.
 
+    // When quota is enforced, this holds the byte ceiling so the increment UPDATE below can
+    // enforce it atomically (`WHERE total + size <= limit`). Without that guard, two
+    // concurrent uploads from the same user (e.g. phone + laptop) both pass this stale
+    // pre-check and both increment, blowing past the quota. The pre-check stays as a
+    // fast path that avoids the disk write when the user is already clearly over.
     let mut quota_limit: Option<i64> = None;
     if quota_on && storage_quota_on {
         let estimate = compute_storage_quota(&state).await;
@@ -1370,7 +1389,9 @@ pub async fn get_thumbnail(
 
 #[cfg(test)]
 mod tests {
-    use super::{MIN_QUOTA_LIMIT_BYTES, RangeSpec, parse_range, quota_limit_bytes};
+    use super::{
+        DISK_RESERVE_BYTES, MIN_QUOTA_LIMIT_BYTES, RangeSpec, parse_range, quota_limit_bytes,
+    };
 
     // `Range` handling exists because iOS Safari probes every `<video>` with
     // `Range: bytes=0-1` and abandons the load without a 206. These pin the forms a
@@ -1523,6 +1544,67 @@ mod tests {
         );
     }
 
+    /// THE INVARIANT THE UPLOAD GATE EXISTS FOR: if an upload is accepted, the keepsake must
+    /// still be buildable afterwards.
+    ///
+    /// The gate and `ensure_export_space` answer the same question at different times, from the
+    /// same `required_free_bytes`. If they ever drift, the failure is silent and terminal — every
+    /// upload succeeds and the archive can never be built, discovered only when the host taps
+    /// release and there is nobody left to fix it. This pins the two together.
+    ///
+    /// Models the real box: 40 GB volume, ~5 GB consumed by OS, images and Postgres.
+    #[test]
+    fn an_accepted_upload_always_leaves_room_to_build_the_keepsake() {
+        const USABLE: i64 = 35 * GB;
+        let reserve = DISK_RESERVE_BYTES;
+
+        // Walk the gallery upward in 250 MB steps and assert the two agree at every point.
+        let mut media: i64 = 0;
+        let step: i64 = 250 * 1024 * 1024;
+        let mut last_accepted = 0i64;
+        while media < USABLE {
+            let free = USABLE - media;
+            let media_after = media + step;
+            let free_after = free - step;
+            let required =
+                crate::services::export::required_free_bytes(media_after as u64, 2) as i64 + reserve;
+            let gate_accepts = free_after >= required;
+
+            if gate_accepts {
+                // The export preflight must agree, using the SAME arithmetic it will run later.
+                let preflight_needs =
+                    crate::services::export::required_free_bytes(media_after as u64, 2) as i64
+                        + reserve;
+                assert!(
+                    free_after >= preflight_needs,
+                    "gate accepted at media={media_after} but the preflight would refuse"
+                );
+                last_accepted = media_after;
+            }
+            media = media_after;
+        }
+
+        // Sanity-check the ceiling is where the arithmetic says: 35 = 2.2·M + 10 ⇒ M ≈ 7.8 GB.
+        // Pinned loosely (6–9 GB) so a deliberate change to the overhead multiplier or the
+        // reserve fails this test loudly rather than silently moving the cliff.
+        assert!(
+            (6 * GB..=9 * GB).contains(&last_accepted),
+            "expected the gallery ceiling near 7.8 GB on a 35 GB volume, got {last_accepted} bytes"
+        );
+    }
+
+    /// The gate must be the binding constraint, not the per-user floor. With 100 guests each
+    /// allowed 500 MB, the per-user quota alone would authorise ~50 GB on a 40 GB disk.
+    #[test]
+    fn the_global_gate_binds_before_the_per_user_floor_can_overfill_the_disk() {
+        let per_user_total = MIN_QUOTA_LIMIT_BYTES * 100;
+        assert!(
+            per_user_total > 35 * GB,
+            "premise: the per-user floor alone over-commits the volume, so the global gate \
+             is what must stop it"
+        );
+    }
+
     /// The floor must never write a cheque the volume cannot cash — otherwise a full disk
     /// still hands out a 500 MB allowance and the filesystem Postgres needs fills up.
     #[test]

backend/src/services/export.rs

@@ -1354,7 +1354,7 @@ const EXPORT_SIZE_OVERHEAD_PCT: u64 = 110;
 /// Computed in `u128` and clamped, NOT with `saturating_mul`: saturating first and then dividing by
 /// 100 quietly turns an overflow into a number ~100x too small, which is the one direction that
 /// matters here — an under-estimate authorises the very write the preflight exists to refuse.
-fn required_free_bytes(media_bytes: u64, armed: i64) -> u64 {
+pub(crate) fn required_free_bytes(media_bytes: u64, armed: i64) -> u64 {
     let needed = media_bytes as u128 * EXPORT_SIZE_OVERHEAD_PCT as u128 / 100
         * armed.max(1).min(i64::from(u32::MAX)) as u128;
     needed.min(u64::MAX as u128) as u64

backend/src/services/media_total.rs

@@ -0,0 +1,82 @@
+//! 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 bytes = sqlx::query_scalar::<_, Option<i64>>(
+            "SELECT SUM(total_upload_bytes)::bigint FROM \"user\"",
+        )
+        .fetch_one(pool)
+        .await
+        .ok()
+        .flatten()
+        .unwrap_or(0)
+        .max(0);
+        *self.inner.write().unwrap() = Some((bytes, Instant::now()));
+        bytes
+    }
+}
+
+impl Default for MediaTotalCache {
+    fn default() -> Self {
+        Self::new()
+    }
+}

backend/src/services/mod.rs

@@ -4,6 +4,7 @@ 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 video;

backend/src/state.rs

@@ -5,6 +5,7 @@ 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;
 
@@ -38,6 +39,8 @@ 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,
 }
 
 impl AppState {
@@ -63,6 +66,7 @@ impl AppState {
             sse_tickets: SseTicketStore::new(),
             config_cache,
             disk_cache: DiskCache::new(),
+            media_total: MediaTotalCache::new(),
         }
     }
 }

docker-compose.yml

@@ -31,14 +31,16 @@ services:
     deploy:
       resources:
         limits:
-          # 1G, not 512M. DATABASE_MAX_CONNECTIONS defaults to 30 for a ~100-guest event
+          # 1G, not 512M. Postgres 16's default shared_buffers plus a pool of backends
-          # (feed polling + SSE + uploads at once), and 30 backends plus Postgres 16's
+          # leaves very little headroom at 512M, and an OOM here does not degrade one
-          # default shared_buffers leaves very little headroom at 512M. An OOM here does
+          # feature — it takes the event down, because every request path touches the
-          # not degrade one feature — it takes the event down, because every request
+          # database.
-          # path touches the database. Memory is the cheaper knob than shrinking the
-          # pool back and reintroducing the queueing it was raised to fix.
           #
-          # Raising DATABASE_MAX_CONNECTIONS further means raising this too.
+          # `.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:

frontend/src/app.html

@@ -46,6 +46,33 @@
 						document.head.appendChild(s);
 					}
 				} catch (_) {}
+
+				// Boot backstop. With `ssr = false` the page is EMPTY until the bundle mounts, so
+				// anything that stops it mounting leaves the guest on the spinner below forever:
+				// a chunk 404 after a redeploy, a dead uplink, or untranspiled syntax on an old
+				// phone. There is no message, no reload control, and in a standalone PWA no URL
+				// bar to escape from — the app is simply bricked for that guest, all evening.
+				//
+				// A TIMEOUT, deliberately, not feature detection: a SyntaxError in the bundle is
+				// invisible to any capability check, whereas "still not painted" catches every
+				// cause at once. The root layout removes #app-boot on mount, so its continued
+				// presence IS the failure signal and nothing needs to cancel this.
+				//
+				// 15s is well beyond a slow-but-healthy load on venue wifi (the bundle is ~130 kB
+				// gzipped); erring long matters more than erring short, because a false positive
+				// here would tell a guest something is broken while it is quietly working.
+				setTimeout(function () {
+					var boot = document.getElementById('app-boot');
+					if (!boot) return; // app mounted — nothing to do
+					boot.innerHTML =
+						'<div style="max-width:20rem;text-align:center">' +
+						'<p style="font-family:Georgia,serif;font-size:1.125rem;font-weight:600;margin:0 0 .5rem">Die App konnte nicht geladen werden</p>' +
+						'<p style="margin:0 0 1.25rem;color:#545350;line-height:1.5">Bitte prüf deine Verbindung und lade die Seite neu.</p>' +
+						'<button id="app-boot-reload" style="font:inherit;font-weight:600;cursor:pointer;border:0;border-radius:.5rem;padding:.625rem 1.25rem;background:#8a6a2b;color:#fff">Neu laden</button>' +
+						'</div>';
+					var btn = document.getElementById('app-boot-reload');
+					if (btn) btn.addEventListener('click', function () { location.reload(); });
+				}, 15000);
 			})();
 		</script>
 		%sveltekit.head%
@@ -65,6 +92,14 @@
 			<span class="app-boot__spinner"></span>
 			<span class="app-boot__label">EventSnap</span>
 		</div>
+		<!-- The app is client-rendered end to end, so with JS off there is nothing at all to
+		     show. Say so, rather than leaving a permanent spinner over a blank page. -->
+		<noscript>
+			<div class="app-boot__noscript">
+				<p><strong>EventSnap braucht JavaScript</strong></p>
+				<p>Bitte aktiviere JavaScript im Browser und lade die Seite neu.</p>
+			</div>
+		</noscript>
 		<style>
 			#app-boot {
 				position: fixed;
@@ -103,6 +138,26 @@
 					transform: rotate(360deg);
 				}
 			}
+			/* Sits above #app-boot, which is `position: fixed` and would otherwise cover it. */
+			.app-boot__noscript {
+				position: fixed;
+				inset: 0;
+				z-index: 1;
+				display: flex;
+				flex-direction: column;
+				align-items: center;
+				justify-content: center;
+				gap: 0.25rem;
+				padding: 2rem;
+				text-align: center;
+				background: #faf9f7;
+				font-family: system-ui, -apple-system, 'Segoe UI', Roboto, sans-serif;
+				color: #545350;
+			}
+			html.dark .app-boot__noscript {
+				background: #100f0f;
+				color: #a6a4a1;
+			}
 		</style>
 	</body>
 </html>

frontend/src/lib/upload-queue.test.ts

@@ -3,7 +3,8 @@ import {
 	classifyUploadStatus,
 	isReversibleLock,
 	entryToQueueItem,
-	shouldAbortForStall
+	shouldAbortForStall,
+	suspendedSinceLastTick
 } from './upload-queue';
 
 /**
@@ -140,4 +141,55 @@ describe('shouldAbortForStall', () => {
 		expect(shouldAbortForStall(now - 91_000, now, true)).toBe(false);
 		expect(shouldAbortForStall(now - 121_000, now, true)).toBe(true);
 	});
+
+});
+
+/**
+ * The watchdog measures SILENCE via `Date.now()`, but a backgrounded phone freezes the
+ * interval while the clock keeps running. Without crediting the un-run window back, the first
+ * tick after a screen lock reads the whole sleep as a stall and aborts a healthy upload —
+ * re-sending from byte zero and spending one of five permanent auto-attempts. That is what
+ * every phone does between shots at a party.
+ *
+ * Detecting the freeze from the tick gap (rather than from `visibilitychange`) also covers the
+ * causes that fire no visibility event at all: a throttled-but-visible tab, a closed lid, an
+ * occluded window.
+ */
+describe('suspendedSinceLastTick', () => {
+	const now = 1_000_000;
+
+	it('credits nothing for a tick that arrived on schedule', () => {
+		expect(suspendedSinceLastTick(now - 5_000, now, 5_000)).toBe(0);
+	});
+
+	it('credits nothing for ordinary timer jitter or throttling', () => {
+		expect(suspendedSinceLastTick(now - 6_900, now, 5_000)).toBe(0);
+	});
+
+	it('credits the whole frozen window when the interval did not run', () => {
+		// Screen locked ~2 minutes: a 5s interval arriving 130s late.
+		expect(suspendedSinceLastTick(now - 130_000, now, 5_000)).toBe(125_000);
+	});
+
+	it('credits nothing when the clock jumps backwards (NTP correction)', () => {
+		expect(suspendedSinceLastTick(now + 60_000, now, 5_000)).toBe(0);
+	});
+
+	it('a suspension longer than the stall ceiling does not abort a healthy upload', () => {
+		// The bug, end to end: 3 minutes suspended, interval resumes, no bytes since.
+		const lastProgressAt = now - 180_000;
+		const credited = Math.min(
+			now,
+			lastProgressAt + suspendedSinceLastTick(now - 185_000, now, 5_000)
+		);
+		expect(shouldAbortForStall(credited, now, false)).toBe(false);
+	});
+
+	it('but a socket still silent 91s AFTER resume is aborted, never left to xhr.timeout', () => {
+		// iOS reaps backgrounded sockets without firing `error`. The credit buys one fresh
+		// window, not immunity — otherwise a dead upload would hang for 5-60 minutes holding
+		// the queue's `processing` latch.
+		const resumedAt = now - 91_000;
+		expect(shouldAbortForStall(resumedAt, now, false)).toBe(true);
+	});
 });

frontend/src/lib/upload-queue.ts

@@ -94,6 +94,45 @@ export function shouldAbortForStall(
 	return now - lastActivityAt > ceiling;
 }
 
+/**
+ * Normal timer jitter/throttling budget. A tick later than `interval + this` did not run
+ * because the page was suspended, not because it was merely late.
+ */
+const SUSPEND_TOLERANCE_MS = 2_000;
+
+/**
+ * Wall-clock the watchdog interval FAILED to cover because the page was suspended.
+ *
+ * `Date.now()` keeps advancing while a backgrounded phone is frozen, but `setInterval` does
+ * not run. So the first tick after a screen lock saw the entire sleep as "no bytes moved" and
+ * aborted a connection that was very possibly healthy — restarting a 200 MB video from byte
+ * zero, burning one of five PERMANENT auto-attempts (`chargeAttempt`), and breaking the drain
+ * loop for every other queued photo. A phone in a pocket between shots is the common case at a
+ * party, not an edge case.
+ *
+ * The interval is its own suspension detector: a tick scheduled 5s out that arrives 130s late
+ * means the page was frozen for ~125s, and that is exactly the window the watchdog had no
+ * right to measure. Deliberately chosen over a `visibilitychange` listener, which only covers
+ * the causes that happen to fire that event — a throttled-but-visible tab, a closed laptop lid
+ * and an occluded window all freeze timers without one. It also needs no listener, no
+ * module-level state, no SSR guard and no teardown.
+ *
+ * `performance.now()` was rejected as the clock source: Safari pauses it across system sleep
+ * on some paths while Chrome does not, which is precisely the non-uniformity that makes it
+ * unusable as the sole signal.
+ *
+ * Returns 0 for a normal tick, and 0 if the clock jumps BACKWARDS (an NTP correction) — that
+ * fails open, and the wall-clock `xhr.timeout` still bounds the request.
+ */
+export function suspendedSinceLastTick(
+	lastTickAt: number,
+	now: number,
+	intervalMs: number = STALL_CHECK_INTERVAL_MS
+): number {
+	const overshoot = now - lastTickAt - intervalMs;
+	return overshoot > SUSPEND_TOLERANCE_MS ? overshoot : 0;
+}
+
 /**
  * Wall-clock cap for one attempt, scaled by file size assuming a floor of ~8 kB/s — a
  * deliberately pessimistic rate, because killing a slow-but-progressing upload would lose
@@ -905,11 +944,25 @@ async function uploadItem(id: string): Promise<void> {
 			// connection that never errors and never completes. Only "no bytes moved" catches
 			// that without also punishing a healthy slow link.
 			let lastProgressAt = Date.now();
+			let lastTickAt = Date.now();
 			let bodySent = false;
 			let stalled = false;
 			const stallTimer = setInterval(() => {
-				if (!shouldAbortForStall(lastProgressAt, Date.now(), bodySent)) return;
+				// Never fire twice. `xhr.abort()` on a request already in `readyState === DONE`
+				// emits NO `abort` event, so `settle()` would never run: the interval would keep
+				// running forever, re-aborting every 5s, and `activeUploads` would keep a stale
+				// entry so the guest's ✕ button silently did nothing.
+				if (stalled) return;
+				const now = Date.now();
+				// Credit back the window the page was frozen. The watchdog measures SILENCE, and
+				// a period in which it could not observe anything is not evidence of silence.
+				// Clamped to `now` so a progress event delivered right at resume cannot push the
+				// timestamp into the future.
+				lastProgressAt = Math.min(now, lastProgressAt + suspendedSinceLastTick(lastTickAt, now));
+				lastTickAt = now;
+				if (!shouldAbortForStall(lastProgressAt, now, bodySent)) return;
 				stalled = true;
+				clearInterval(stallTimer);
 				xhr.abort();
 			}, STALL_CHECK_INTERVAL_MS);
 			const settle = (fn: () => void) => {
@@ -1018,7 +1071,16 @@ async function uploadItem(id: string): Promise<void> {
 					else reject(new NetworkError('Abgebrochen'));
 				})
 			);
+			// `send` can throw SYNCHRONOUSLY — most plausibly on a phone whose OS purged the
+			// backing store for the blob, leaving a neutered File. The executor would turn that
+			// into a rejection and `uploadItem` would recover, but `settle()` never runs: the
+			// stall interval leaks and `activeUploads` keeps a stale entry, so the ✕ button on
+			// that item stops working for the rest of the session.
+			try {
 				xhr.send(formData);
+			} catch {
+				settle(() => reject(new NetworkError('Netzwerkfehler')));
+			}
 		});
 
 		// Success — remove blob from IndexedDB, mark done

frontend/src/tailwind-theme.css

@@ -95,11 +95,81 @@
 	--color-purple-700: #6f5523;
 	--color-purple-800: #59441e;
 	--color-purple-900: #493819;
+	/* The ramp stopped at 900 while blue/primary both run to 950, so
+	 * `dark:bg-purple-950/50` (routes/host/+page.svelte) fell through to Tailwind's default
+	 * violet — off-brand on every browser, modern ones included. Mirrors `--color-blue-950`. */
+	--color-purple-950: #29200d;
 	--color-violet-500: #ab8433;
 	--color-violet-600: #8a6a2b;
 	--color-accent-500: #ab8433;
 	--color-accent-600: #8a6a2b;
 
+	/* ── Semantic: red / amber / green ───────────────────────────────────────────
+	 * PINNED TO HEX rather than inherited. Tailwind v4 ships these families as
+	 * `oklch()`, which Safari <15.4, Chrome <111 and Samsung Internet <22 (the default
+	 * browser on Samsung phones) cannot parse. The declaration is accepted but
+	 * `var(--color-red-600)` is then invalid at computed-value time, so `background-color`
+	 * falls back to `initial` — transparent — and `.btn-danger` in the component layer
+	 * renders white text on nothing. An invisible delete-confirm button.
+	 *
+	 * Note this is NOT the `@apply` hard-fail described for `primary` above: these families
+	 * have Tailwind defaults, so an unpinned stop does not break the build, it silently
+	 * reverts to oklch. Full 50-950 ramps are therefore about closing that silent-leak class
+	 * permanently, not about compiling.
+	 *
+	 * Values are Tailwind 4.2.2's own defaults gamut-mapped to sRGB by Lightning CSS — the
+	 * same converter already in this build pipeline, which is why they match the hex
+	 * fallbacks it emits for the `/alpha` opacity forms (verified against `#bf000f`,
+	 * `#460809`, `#461901`, `#032e15`, `#82181a`, `#ffa3a3`, `#0d542b` in the shipped CSS).
+	 * Naive channel clipping gives different, wrong values for the out-of-gamut stops.
+	 * Modern browsers therefore render exactly what they render today. */
+	--color-red-50: #fef2f2;
+	--color-red-100: #ffe2e2;
+	--color-red-200: #ffcaca;
+	--color-red-300: #ffa3a3;
+	--color-red-400: #ff6568;
+	--color-red-500: #fb2c36;
+	--color-red-600: #e40014;
+	--color-red-700: #bf000f;
+	--color-red-800: #9f0712;
+	--color-red-900: #82181a;
+	--color-red-950: #460809;
+
+	--color-amber-50: #fffbeb;
+	--color-amber-100: #fef3c6;
+	--color-amber-200: #fee685;
+	--color-amber-300: #ffd236;
+	--color-amber-400: #fcbb00;
+	--color-amber-500: #f99c00;
+	--color-amber-600: #dd7400;
+	--color-amber-700: #b75000;
+	--color-amber-800: #953d00;
+	--color-amber-900: #7b3306;
+	--color-amber-950: #461901;
+
+	--color-green-50: #f0fdf4;
+	--color-green-100: #dcfce7;
+	--color-green-200: #b9f8cf;
+	--color-green-300: #7bf1a8;
+	--color-green-400: #05df72;
+	--color-green-500: #00c758;
+	--color-green-600: #00a544;
+	--color-green-700: #008138;
+	--color-green-800: #016630;
+	--color-green-900: #0d542b;
+	--color-green-950: #032e15;
+
+	/* Avatar chips (lib/avatar.ts) — same oklch problem, same fix. Only the stops those
+	 * chips actually use; nothing else in the app references rose or teal. */
+	--color-rose-100: #ffe4e6;
+	--color-rose-200: #ffccd3;
+	--color-rose-700: #c20039;
+	--color-rose-900: #8b0836;
+	--color-teal-100: #cbfbf1;
+	--color-teal-200: #96f7e4;
+	--color-teal-700: #00776e;
+	--color-teal-900: #0b4f4a;
+
 	/* ── Neutrals: pearl → silver → graphite (remaps `gray-*`). Whisper-warm
 	 * pearl at the light end (research: "warm pearl, not stark white"), turning
 	 * neutral-cool through the mids/darks so structure reads as silver, not sand. */