On Fri, Apr 5, 2024 at 4:20 AM Melanie Plageman
<melanieplageman(at)gmail(dot)com> wrote:
> So, sequential scan does not have per-buffer data. I did some logging
> and the reason most fully-in-SB sequential scans don't use the fast
> path is because read_stream->pending_read_nblocks is always 0.
Hnghghghgh... right, sorry I guessed the wrong reason, it turns out
that I made a fast path just a little too specialised for pg_prewarm.
Working on it...