Updated with FB24122131. Here's the ready-to-paste forum post:
Title: registerForRemoteNotifications() never completes (no success, no error callback) since July 3 — App Store Connect metrics confirm same cutoff
Body:
Posting in case anyone else has hit this — happy to compare notes, and I've also filed FB24122131 with Apple directly.
Since ~July 3, 2026, our app (MyRoost, iOS) has been completely unable to register for remote notifications. registerForRemoteNotifications() is called at launch, but neither didRegisterForRemoteNotificationsWithDeviceToken: nor didFailToRegisterForRemoteNotificationsWithError: ever fires. No token, no error — just silence, indefinitely.
This isn't a new-integration bug — the same code worked fine through July 2. What's interesting is that App Store Connect's own "Push Notifications → Overview" dashboard for our app independently shows notifications received by APNs dropping from ~20/day to flat zero starting exactly July 3, matching our own logs precisely.
We've ruled out essentially everything on our side:
- Push Notifications capability + entitlements (independently verified present in the actual signed binary, not just source, via a CI step that dumps the compiled .ipa's Info.plist and codesign entitlements)
- Firebase/APNs Authentication Key configuration
- OS-level notification permission (confirmed granted)
- MDM/configuration profiles (none), Screen Time restrictions (none)
- Wi-Fi vs. cellular (identical failure on both)
- Full device reboot, clean reinstall, account logout/login
- Reproduced identically on two separate physical devices (iPhone 17 Pro, iOS 26.5.2; iPad A16, iPadOS 26.5.2)
- Added native diagnostic logging to capture the raw error from
didFailToRegisterForRemoteNotificationsWithError:directly — confirms the native callback itself never fires, not just that we're failing to observe it
Filed an Apple Developer Support case (102936910825) on July 8 — after investigation, closed 2026-07-30 as "outside scope" and redirected here / to Feedback Assistant.
Has anyone else seen a total, silent APNs registration failure like this, starting around the same early-July timeframe? Any workaround, or confirmation this is a known platform-side issue, would be hugely appreciated.