As a follow-up, we isolated the affected subsystem under its own supervisor so that any similar external change or failure will be contained without affecting regional service availability.

A fix has been deployed to the EU region. Wildcard certificates now verify correctly and application instances are starting normally. We are monitoring.

The issue has been identified: Firebase (FCM) rotated the TLS certificate on their endpoint to a wildcard certificate, and our TLS handshake failed to verify it. The failed connections caused application instances in the EU region to crash on startup.

We are investigating a service degradation in the EU region.