Fix 'unreachable code' panic in Persistent Subscriptions

[ryanw]

I encountered this unreachable!() panic quite a lot when using Persistent Subscriptions. I'm not entirely sure the cause, but as I am hitting it then it's fair to say it's not "unreachable". By changing this to an Err it resolves my panic issue.

[CLAassistant]

All committers have signed the CLA.

[YoEight]

Hey @ryanw,

Thanks for your contribution. The reason why that unreachable was there is because it shouldn't be reachable, thus there is a bug that needs fixing or at least we need to make sure if the motivation that lead to that decision is no longer valid.

I assume you are experiencing that issue quite reliably, would you mind sharing logs from the client or from the server? That would help narrow down the cause of what is happening here.

Also what version of the server are you using? What kind of topology are we talking about ? (single node or cluster).

I appreciate it might not be possible but would you be able to provide a reproducible snippet of your code causing that bug?

Thanks for your time.

[YoEight]

Cluster tests are failing because the old docker-compose script is no longer available in ubuntu-latest. I will provide a patch tomorrow to address it.