Usually we aren't deploying a schema change that's really huge so we just go for it and let the application crash for those users who happen to hit a place wher

Usually we aren't deploying a schema change that's really huge so we just go for it and let the application crash for those users who happen to hit a place where the code/schema are out of sync. Zero (no crash) downtime deployments seems like too much effort for too little gain.