It would be great if we could monitor the status of a marathon deployment before exiting. Ideally, we could even set a timeout (with a sensible default) which would trigger an automatic rollback.