List the subjects a rule covers
What It Does
A caller with population and dataset query authority requests subjects qualifying under a release and day, with optional scope, comparison or export.
The Flow
Prepare the release and plan, check supported providers and subject dataset, bound named subjects at 250,000, create run and job atomically, then start after commit. Execute the durable job’s evaluation, completion and failure steps. Read ordered dataset pages, prune permitted database-only exclusions, evaluate remaining subjects in batches of one hundred and atomically store counts, subjects and cursor. Calculate comparison deltas, produce any requested export, set the intended ninety-day expiry and finish a dependent simulation. Read the submitted run and cursor pages of retained subjects using its returned identity.
When Something Fails
Preparation is refused before insertion; a failed post-commit start leaves the job for reconciliation. A submitted run can remain queued after a start failure. Failed or timed-out executions settle the job and trigger alarms. Run and dependent simulation failure remain visible. Retries recompute a page but cannot commit it twice; an unallowed million-subject limit fails the run. Previously committed page results remain on a failed run. Completion failure marks the run and simulation failed. An export written before failure can remain without cleanup. Unknown or other-workspace identities return not found. Lookup makes no business changes.
Open Design Questions
Without pruning, subjects excluded by the database-only prefix still contribute ineligible counts.