Re: BUG #17245: Index corruption involving deduplicated entries

On Thu, Oct 28, 2021 at 2:40 PM Pavel Borisov <pashkin.elfe@gmail.com> wrote:
> One hypothesis just out-of-the air:
> If heap page visibility map bit is wrongly set to all_visible, index-only scan can give duplicates (in effect new and
oldversions of the same row), with no apparent corruption of neither index nor heap relation.
 

That can't be it, because affected physical index pages directly
contain duplicate heap TIDs (and plenty of them). That's never allowed
to happen, no matter what.

-- 
Peter Geoghegan