Meta was launching Threads at product speed — fast public rollout, global scale from day one, dual-platform delivery. They needed QA coverage across iOS, Android, and web in a compressed window with zero room for critical failures at GA. The surface area was large; the timeline was not.
We ran three parallel testing tracks over 14 days across the full surface:
On iOS 16, a provisional notification permission state caused all push delivery to fail without surfacing an error. Affected ~18% of test devices in that OS bracket.
Composing a thread, backgrounding the app, then rotating the device before returning destroyed the draft without recovery. Reproducible 100% on both platforms.
Password reset response time differed measurably (~240ms) between registered and unregistered email addresses, enabling account enumeration at scale.
Thread reply indentation collapsed at nesting depth 4+, making long conversations visually flat and unreadable on Android. Not reproducible on iOS.
Avatar image load requests were URL-encoding usernames inconsistently, causing 404s for accounts with Arabic, CJK, or Cyrillic characters in their handle.
Returning to the feed from a post detail view always reset scroll position to top, making it impossible to resume browsing from where the user left off.
Double-tapping to like a post triggered two distinct haptic pulses ~80ms apart instead of one. Minor but noticeably off — reported by 3 of 5 user testers independently.
Clearing all notifications via "Mark all read" left the app badge at 1 indefinitely until next cold launch. The OS badge count was correct; the in-app counter was not.
Focus was not constrained inside modal dialogs, allowing Tab to navigate behind the overlay. Fails WCAG 2.1 SC 2.1.2.
A 200–400ms flash of the empty state appeared on cold launch before content loaded, even on fast connections. Felt like a broken state to test users.
In threads containing horizontally scrollable media carousels, the iOS swipe-back gesture triggered ~40% of the time when the intent was to scroll media.
Before our review cycle ended, the team shipped fixes for all 3 critical issues and 6 of the 8 remaining high-severity findings. Zero critical issues surfaced in public launch reviews on day one.
The 2 remaining medium findings were tracked and resolved in the first post-launch sprint. The cross-platform delta report became a reference document for the team's ongoing QA process.