Occasionally jobs hits the Runner's timeout so I think it's better to automatically retry them in those cases since there's not much we can do at the application level.