We're upgrading our primary Postgres instance type. New requests might fail. All received requests & agent tasks are queued up and will progress once the database is up and running again.