PostgreSQL's fsync() surprise

I'm not sure I agree that it's really worse. Offering different behavior based on whether the inode got evicted from the cache is pretty nasty, as userland has no way to detect whether that has happened.

That said, I'm not opposed to re-enabling the ability to see unreported errors that occurred prior to the open (and your scheme to do that was pretty clever) if the Pg folks think it's of value in the near term. Maybe we could make that behavior opt-in based on a sysctl or something?