Comment 5 for bug 31651

I think this, bug 32606 and bug 32615 may all be the same problem.

I think flush_database_updates() is a red herring here. It isn't a case of sqlobject buffering up changes: the SQL statement log in these OOPS reports clearly shows that the commands to create these table rows are being sent to the database. The question is why they don't get returned by the subsequent select queries.