LockReplace: xid table corrupted

Hello all,

Does anyone know what this message in the postmaster's log means?

1010125.09:08:15.323 [415] NOTICE: LockReplace: xid table corrupted

We have a Postgres (6.5.3) database running on a Linux (2.2.13) server.

In our database we have a main table 'pallets' which currently has about 150,000
rows. We vacuum
twice daily - at 9:05 am and then again at 9:05 pm. This database is replicated
to another identical box
at the application code level. What we have noticed is that now and again we
lose data from the
database! We know this because the data is in the spare database but not the
master. The data seems to
be in the database for a while after it was first written (we know because
transactions to higher-level systems
show the data) but then at some time after it just disappears!

We are wondering if it is something to do with vacuuming? We obviously don't
lose data everytime we vacuum
but what if we see the above error message while vacuuming? Could that mean
we've lost data?

At present running vacuum verbose analyze of the db takes approx 4 minutes 20
seconds.

Any help/other people's experience with this would be greatly appreciated. We
are desperate.

TIA

Les Halliday
Control Engineer
Pedigree Masterfoods