Address remaining panic-on-persistence failures

[tnull]

In #374 we fixed the panics in event handling, but there are still some others that are unaddressed. We still need to fix them, which is important for any kind of remote persistence, especially VSS.

Note that many of the -on-persistence-failure-panics are happening in upstream code or are blocked by upstream developments. So this is marked 'blocked on upstream' for now.