As the containers get restarted whenever we deploy a new version, we should make sure that Postgres gets closed gracefully. This might be the standard approach, but we should check the docs; and then fix it if it *isn't* the default approach.