Some oversized assessor packets fail in a batch after chunking. How should failures be recovered?
Select an answer to reveal the explanation.
Short Explanation
When a fat packet fails, chop it and resend just that ticket by custom_id—like remailing one lost package. Targeted resubmission beats rerunning everything.
Full Explanation
When oversized assessor packets fail inside a batch after chunking attempts, recovery should resubmit failed documents by custom_id rather than replaying the entire successful cohort. custom_id exists so municipal operators can target retries, correlate results, and avoid duplicate charges of compute on packets that already succeeded. After chunking an oversized packet into processable pieces, resubmit only the failed identifiers and reconcile outputs. Resubmitting the entire successful batch without using identifiers wastes cost and risks duplicate downstream writes if consumers are not perfectly idempotent. Dropping custom_id so failed documents cannot be targeted removes the handle needed for surgical recovery. Ignoring failures and marking every packet complete falsifies extract completeness for property records and invites silent gaps in assessment data. Exam caveat: chunking changes document segmentation—ensure reconstituted extracts still satisfy schema and page-conflict rules after targeted resubmit. Operational check: maintain a failure ledger keyed by custom_id, chunk and resubmit only those IDs, and confirm successful IDs are not recharged while failed IDs eventually reach a terminal success or human-escalation state.