We've identified this problem, and believe negative caching of DNS records to be all that remains. We are going to continue to check for at least thirty minutes, the longest time reported for negative caching so far.

Investigating - There is a defect in Postgres that causes unbound DNS records. Negative caching can exacerbate the incident after our fix. Databases affected: many at the outset, some now. Diagnosing the last records.

We're still trying to understand the root cause of the issue at the moment, this is not affecting all managed databases, only a handful of them.