In an attempt to get a hackfix for http://pgfoundry.org/tracker/index.php?func=detail&aid=1011203&group_id=1000411&atid=1376, I'm wonder if it's true that, when looking at pg_locks, the only pid which will have virtualxid = '1/1' and virtualtransaction = '-1/0' will be the bgwriter. That seems correct on all my dbs, but that's hardly a definitive test.