"Curtis Faith" <curtis@galtcapital.com> writes:
> I don't dispute their conclusions in that context and under the
> circumstances they outline of random distribution of deletion and
> insertion values for the index keys. [But the random-distribution
> assumption doesn't always hold.]
That's a fair point. Nonetheless, we have little choice: we cannot
move keys around during concurrent operations. If we push keys to
the right, we may cause an indexscan moving left to miss them, and
vice versa. So we can only eliminate empty pages.
We could possibly allow VACUUM FULL to collapse partly-full pages,
since in that case we know there are no concurrent scans.
regards, tom lane