Read Task

Retry all failed jobs in a batch. This endpoint resets all jobs with state='error' back to state='pending', clears their output data, increments their retry count, and re-dispatches the first pending job of each affected task to the SQS queue. Failed jobs whose input payload the retention sweep already cleared are skipped and counted in `jobs_skipped_expired` — retrying them would re-dispatch an empty payload. Retention is irreversible, so those jobs stay in `error`. No quota is consumed for retries - the retry_count field distinguishes retry attempts from new job executions. Args: batch_id: The batch ID of the batch to retry Returns: Response with retry statistics including jobs reset and tasks affected Raises: 404: If batch not found 403: If user lacks permissions to retry this batch