"Paul Laughlin" <plaughlin(at)gmail(dot)com> writes:
> We got it early last week and again this morning. Before these two it was
> about six months ago.
A certain amount of clustering could be expected, if a lot of the
entries were made at the time of initial table load --- they'd have
nearby OIDs. You can either ignore it (I doubt you need REINDEX, just
retry the update) or else update to PG 8.1 ...
regards, tom lane