Comment 1 for bug 1062324

For the record, the solution was to choose a key on the _new_ table for the DELETE trigger. I asked everyone at Percona and no one seemed to think this would have weird side-effects, nor did I, because we're still deleting rows based on a unique key.