We are currently unsure whether critical failures to publishing individual batches really breaks function execution. We need to ensure that if one batch fails, the whole function fails and retries.
