Re: Bug in Physical Replication Slots (at least 9.5)?

Poking this.  Looking back through the discussion, this issue has been reproduced by multiple people.  The patch still applies to HEAD without issues.  I have no experience with PostgreSQL replication, so I'm not qualified to really review this.  From what I can see with the patch, it's just a small block of code added to /src/backend/replication/walreceiver.c to handle some edge case where the WAL file no longer exists or something.

I think one thing that would help move this forward is if we edited the patch to include a comment explaining why this new code is necessary.  There's lots of great discussion on this issue in the email list, so if a summary of that gets into the code I think it would make the patch easier to understand and make the new walreceiver.c less confusing.