For about 4 minutes starting at 06:39am UTC a significant number of API-requests were rejected.

The root cause was an automatic database fail-over that some servers didn't pick up correctly, which meant they kept querying the old database which caused the 500 error codes.

Timeline:

At 04:25am UTC an automatic database failover was triggered.
At 08:55am UTC we identified the cause of the issue and deployed a fix.
At 09:06am UTC the fix was deployed successfully and we kept monitoring the results.
At 09:10am UTC the issue was resolved.