"Thomas F. O'Connell" <tfo@sitening.com> writes:
> does pg_autovacuum as currently written in contrib vacuum temp
> tables, and, in 8.0, is this then able (however unlikely) to cause
> the sort of error I encountered yesterday?
No, and no, and still no for the integrated version. The
isOtherTempNamespace() checks in vacuum.c and analyze.c absolutely
guarantee this, whether or not the autovacuum code itself is bright
enough to exclude temp tables.
regards, tom lane