While we do not expect any outage, there may be a small window where requests may fail as services are switched over. In the event we observe a large spike or continued errors after the deploy, we may roll back the changes to the current state to avoid service disruption.