>>And most databases get a mix of updates and selects. I would expect it would
>>be pretty hard to go that long with any significant level of update activity
>>and no vacuums and not notice the performance problems from the dead tuples.
>
>
> I think the people who've managed to shoot themselves in the foot this
> way are those who decided to "optimize" their cron jobs to only vacuum
> their user tables, and forgot about the system catalogs.
That's certainly the case with one of the people we helped in IRC - 3
user tables only being vacuumed.
Chris