> We don't expect unrecognized option here and for such a thing, we use
> elog in the code. See the similar usage in
> parseCreateReplSlotOptions().
"pgoutput" is useful for a lot of applications other than our logical
replication subscriber. I think we should expect anything and handle
errors nicely.
> I think we should move to 0002 patch now. In that, I suggest preparing
> separate back branch patches.
They are attached.