Jeff Frost <jeff@frostconsultingllc.com> writes:
> A look at the code revealed that the temp tables were created via normal
> methods, so I'm curious to know if there is a bug regarding temp tables not
> going away on Postgresql-8.2.x after the connection is closed?
You aren't the first to report such a thing, but nobody has the foggiest
idea how it could happen short of a backend crash. Have they had any
crashes lately (or more specifically, around the mod times of those
files, if you checked them)?
Also, were the pg_temp schemas you zapped particularly high-numbered?
Low-numbered ones would get cleaned out on the next use, but if the
crash happened at a peak in the number of active backends it's easy
to believe the files might hang around for awhile.
regards, tom lane