A 311 job must keep human quarantine review separate from automatic gold publish. How should tasks be structured?

Separating quality-review tasks from automatic publish tasks gives civic jobs human control over suspect data. A 311 job should keep quarantine review as its own task (or path) while clean rows follow an automatic gold publish task. That structure lets supervisors adjudicate quarantined tickets without blocking the clean publish path and without gluing review and publish into one irreversible button.

Separating the quarantine review task from the automatic gold publish task with task values mistakes data sharing for structural separation: task values only pass small outputs between tasks, so they do not create a separate human adjudication path and would leave review coupled to the publish it must stay separate from instead of adjudicating quarantined rows on its own task path. Separating the quarantine review task from the automatic gold publish task via repair rerun confuses retry with review: repair rerun only re-executes a failed publish after a fix, it does not create a separate human adjudication path, so quarantined tickets would be handled by rerunning the automatic publish instead of staying pending review while clean rows publish. Separating the quarantine review task from the automatic gold publish task with one run-as confuses identity with separation: one run-as only assigns a single execution identity, it does not create distinct human versus automatic task paths, so review and publish would share the same permissions instead of keeping human control separate from the automatic gold publish.

Exam caveat: separation implies SLA for review backlog—unreviewed quarantine can grow without alerting and aging policies. Operational check: land a mixed clean/bad 311 batch, confirm clean rows auto-publish to gold, confirm bad rows remain in quarantine pending review, complete a review task that releases or rejects samples, and verify gold never received unreviewed quarantine rows.