Craig Ringer <craig(at)postnewspapers(dot)com(dot)au> writes:
> Tom Lane wrote:
>> I'm betting on varying degrees of table bloat. Have you tried vacuum
>> full, cluster, etc?
> Or, if you have been using VACUUM FULL, try REINDEXing the tables,
> because it could easily be index bloat. Clustering the table will take
> care of index bloat as well as table bloat.
Index bloat wouldn't explain the slow-seqscan behavior the OP was
complaining of. Still, you're right that if the tables are bloated
then their indexes probably are too ... and that VACUUM FULL alone
will not fix that.
regards, tom lane