We have a job every morning that clones the current production db instance and then deploys it onto a staging instance. All PRs with schema changes will be applied to this instance when merged. If something goes wrong, we just redeploy the saved image.