which is its own sort of monitoring/alerting problem, because you'll need to know that your deploy failed to come up