When committing time-series buckets atomically (here and here), the logic is inside of a try/catch block so that upon an exception we can abort the batches with the proper error code. However, currently since we are not catching anything outside of DBException, if there is a different type of exception currently we will erroneously try to abort with an OK status.