test(e2e): fix load-induced mobile flakes (gesture timing + hydration)

Three distinct load-induced flakes in the mobile suite, all root-caused from real traces:

1. gestures-longpress "quick tap": longPress() drove page.mouse.down + Node-side
   waitForTimeout(200) + page.mouse.up — three CDP round-trips. Under load the browser saw
   the pointerup >500ms after pointerdown, firing the app's real long-press so the
   ContextSheet opened on a quick tap. New quickTap() helper dispatches pointerdown/up
   entirely in-browser on one clock, so the earlier-due pointerup always cancels the 500ms
   timer first — immune to CDP latency.

2/3. upload-cancel-confirm and sheet-escape interacted with server-rendered controls before
   Svelte hydrated them (caption fill lost -> cancel() navigates to /feed; custom role=radio
   / leave-button click no-ops). Wait for a post-hydration readiness signal (composer
   auto-focus / real profile name) before the first interaction — the same barrier the
   passing feed-based specs already have. (These are now belt-and-suspenders given ssr=false,
   but keep the specs robust regardless of render mode.)

Verified: mobile 22/22, and 0/50 full-mobile runs under load (was ~3/95 pre-fix).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

e2e/helpers/touch.ts

@@ -34,6 +34,47 @@ export async function longPress(page: Page, locator: Locator, durationMs = 600)
   await page.mouse.up();
 }
 
+/**
+ * A sub-threshold press dispatched ENTIRELY in-browser: `pointerdown`, an in-browser
+ * `setTimeout(holdMs)`, then `pointerup` — all on the element carrying `use:longpress`.
+ *
+ * Why not `longPress` (page.mouse + waitForTimeout) for the "quick tap must NOT long-press"
+ * case: those are three independent CDP round-trips with a Node-side wait between the down
+ * and the up. Under cross-test load the browser can process the `pointerup` far later than
+ * `holdMs` of *browser* time, so the app's real 500 ms long-press timer fires and the
+ * ContextSheet opens on what the test intended as a quick tap. (Observed: a 200 ms-intended
+ * tap opening the sheet ~1 % of full-suite runs.) Dispatching both events from one
+ * `evaluate` puts the down→up gap and the app's 500 ms timer on the SAME clock: even if the
+ * event loop stalls, the earlier-scheduled `pointerup` (holdMs) still fires before the
+ * later 500 ms timer, so the relationship holds. Mirrors `doubleTap`.
+ *
+ * No trusted `click` follows a synthetic `pointerup`, so use this ONLY where the assertion is
+ * that the long-press did NOT fire — not where a real click must land (see `longPress`).
+ */
+export async function quickTap(locator: Locator, holdMs = 200) {
+  await locator.evaluate((el: HTMLElement, ms: number) => {
+    const rect = el.getBoundingClientRect();
+    const x = rect.left + rect.width / 2;
+    const y = rect.top + rect.height / 2;
+    const mk = (type: string) =>
+      new PointerEvent(type, {
+        bubbles: true,
+        cancelable: true,
+        clientX: x,
+        clientY: y,
+        pointerType: 'touch',
+        isPrimary: true,
+      });
+    el.dispatchEvent(mk('pointerdown'));
+    return new Promise<void>((resolve) =>
+      setTimeout(() => {
+        el.dispatchEvent(mk('pointerup'));
+        resolve();
+      }, ms)
+    );
+  }, holdMs);
+}
+
 /**
  * Two rapid pointer-event pairs within the doubletap action's 300 ms window.
  *

e2e/specs/09-mobile/gestures-longpress.spec.ts

@@ -12,7 +12,7 @@
  */
 import { test, expect } from '../../fixtures/test';
 import { uploadRaw } from '../../helpers/upload-client';
-import { longPress } from '../../helpers/touch';
+import { longPress, quickTap } from '../../helpers/touch';
 import { readFileSync } from 'node:fs';
 import { join } from 'node:path';
 
@@ -67,8 +67,12 @@ test.describe('Mobile — long-press gesture', () => {
     const card = page.locator('article').filter({ hasText: g.displayName }).first();
     await expect(card).toBeVisible({ timeout: 10_000 });
 
-    // Simulate a short press (200 ms — well under the 500 ms threshold).
+    // A short press (200 ms — well under the 500 ms threshold), dispatched in-browser so
-    await longPress(page, card, 200);
+    // its duration is measured on the browser's own clock. `longPress` (page.mouse +
+    // Node-side waitForTimeout) can't guarantee that under load: the pointerup can reach
+    // the browser >500 ms after the pointerdown even when Node waited only 200 ms, firing
+    // the long-press and opening the sheet on a "quick tap". See `quickTap`.
+    await quickTap(card, 200);
 
     // Within 1 s, the ContextSheet must not be open (aria-modal is set only when
     // open). A quick tap opens the lightbox instead, which is a different element.

e2e/specs/09-mobile/sheet-escape.spec.ts

@@ -11,6 +11,13 @@ test.describe('Mobile a11y — sheets dismiss on Escape', () => {
     await signIn(page, g);
     await page.goto('/account');
 
+    // /account is server-rendered, so its controls exist and look clickable BEFORE Svelte
+    // hydrates their handlers. The data-mode options are custom `div[role=radio]` (no native
+    // input fallback), so a pre-hydration click does nothing and the sheet never opens (a
+    // reproduced ~1% flake). The profile name/expiry are filled in onMount (post-hydration);
+    // waiting for the real display name is a reliable "the page is interactive now" barrier.
+    await expect(page.getByText(g.displayName)).toBeVisible();
+
     // Click the "Original" radio in the Datennutzung section to open the warning sheet.
     // The radio's accessible name is its title + description ("Original Lädt die
     // Originaldateien…"), so anchor on the start, not the end.
@@ -33,6 +40,10 @@ test.describe('Mobile a11y — sheets dismiss on Escape', () => {
     await signIn(page, g);
     await page.goto('/account');
 
+    // Wait for /account to hydrate before tapping the (server-rendered) leave button, so the
+    // click reaches a live handler. See the test above.
+    await expect(page.getByText(g.displayName)).toBeVisible();
+
     await page.getByRole('button', { name: /Event verlassen/i }).click();
     const sheet = page.getByTestId('confirm-sheet');
     await expect(sheet).toBeVisible();

e2e/specs/09-mobile/upload-cancel-confirm.spec.ts

@@ -19,7 +19,13 @@ test.describe('Mobile — upload composer cancel confirmation', () => {
     await page.goto('/upload');
 
     const caption = page.getByTestId('upload-caption');
-    await expect(caption).toBeVisible();
+    // The composer is server-rendered, so the textarea is visible (and fillable) BEFORE Svelte
+    // hydrates its `bind:value`. A fill that lands pre-hydration sets the DOM value but never
+    // reaches the reactive `caption` state — so `cancel()` sees an empty caption and navigates
+    // straight to /feed instead of opening the discard sheet (a reproduced ~1% flake). The
+    // composer auto-focuses the textarea in onMount (which only runs post-hydration), so waiting
+    // for focus is a reliable "the bindings are live now" barrier before we type.
+    await expect(caption).toBeFocused();
     await caption.fill('a meaningful caption that I do not want to lose');
 
     // Tap the close (X) button in the composer header.
@@ -44,6 +50,9 @@ test.describe('Mobile — upload composer cancel confirmation', () => {
     await signIn(page, g);
     await page.goto('/upload');
 
+    // Wait for the composer to hydrate (onMount auto-focuses the textarea) before tapping
+    // Abbrechen, so the click reaches a hydrated button. See the test above.
+    await expect(page.getByTestId('upload-caption')).toBeFocused();
     await page.getByRole('button', { name: 'Abbrechen' }).click();
     await page.waitForURL('**/feed', { timeout: 3_000 });
   });