We are still monitoring the situation, but as of right now, the issue appears to be resolved.

We've detected a database failure followed by a delayed switchover, which caused issues for a small number of applications.