Multiple reports from users that the os_user_group table is cleared occasionally.

Need to know which database/appserver they're all using, since this doesn't happen on any of our resin+postgres servers.