AW: AW: AW: AW: AW: Vacuum only with 20% old tuples

> > But yes the rename log entries (only those) would need to be 
> > written immediately. Is this a performance issue? I guess not.
> 
> Two fsyncs per rename is probably not a performance killer, but you'd
> have to also fsync log entries for file creates and deletes, which
> might be a bigger deal.

I am not sure, but I don't think those are necessary,
only if you want to guarantee no leftover files on backend abort.
I agree that create/delete would be critical to performance.

Andreas