Between 23:35 and 23:45 UTC on June 9, phone call requests failed due to a database migration conflict during a deployment. A migration operation queued behind a long-running query, causing an exclusive lock that blocked all queries on the calls table for approximately 10 minutes.

The migration was stopped and rolled back cleanly with no data loss. Production remained on the previous image and is now fully operational.

We've implemented a lock timeout safeguard to prevent similar stalls in future deployments.