Ben Chobot <bench(at)silentmedia(dot)com> writes:
> And unfortunately, Tom, we're not resetting stats counters. :(
Mph. Well, the other thing that comes to mind is that n_live_tup
(and n_dead_tup) is typically updated by ANALYZE, but only to an
estimate based on ANALYZE's partial sample of the table. If the
sample isn't very representative then it might be far off.
How does n_live_tup compare to an actual COUNT(*) count?
regards, tom lane