Mike Schanne <mschanne@kns.com> writes:
> Is this what you are referring to?
> - Prevent VACUUM from trying to freeze an old multixact ID involving a still-running transaction (Nathan Bossart,
JeremySchneider)
> This case would lead to VACUUM failing until the old transaction terminates.
> https://www.postgresql.org/docs/release/9.6.16/
Hmmm ... after digging through the commit log, it seems the improvements
I was thinking of were all pre-9.6. The only post-9.6 vacuum truncation
performance fix I can find is
https://git.postgresql.org/gitweb/?p=postgresql.git&a=commitdiff&h=7e26e02ee
which came in in v10.
regards, tom lane