> In my testing it looks like pg_stat_reset was ran, but I haven't been able > to verify this from my logs. > > Is there another way the stats could get reset?
Yes, a crash/restart does that.
> Does autovacuum use the stats from pg_stat_user_table meaning if there 50 > live tuples and 1000 dead tuples, but then the stats are reset, those dead > tuples would never get cleaned up until there was new activity on that > table?