Message 254514 - Python tracker

Thanks, but I don't like the idea of that patch. It feels like a hack that makes it less likely that the issue occurs, but I don't feel we should rely on the callbacks being called before checking the selector again. There may be other reasons (perhaps a future modification to the code) why we might occasionally check the selector redundantly. IOW I think we should really ensure that all I/O callbacks are properly idempotent.