Our Cloud provider rebooted one of our database machines. It should have been transparent for the platform, but there was a bug in the failover mechanism which prevented the recovery. It was manually fixed, and the bug squashed to prevent any future issue.