The problems resolved automatically at 09:41 UTC. We are monitoring the situation.

We're aware that for about 5 minutes from 09:36 to 09:41 UTC, the Replicate API returned HTTP 500 errors for many calls, including particularly for prediction creates. It appears that an essential database failed or was unreachable, and failover mechanisms don't appear to have worked correctly. We are looking into what happened.

All workloads are currently functioning correctly.