workqueue-census.tsv : tab-separated census (one row per Q heading).
workqueue-census.md : markdown table of the same rows.
extract_manifest.json: schema, counts, provenance.
CHECK : independent verifier (re-derives expected rows from WORK-QUEUE.md; does NOT import this extractor).

Conventions

A row is created for every line matching '## ...' where is 'Q' followed by digits (continuation labels such as closure, section 9 follow-on, verified, claimed are kept verbatim in the title).
q : the label 'Q' (dupes kept as separate rows).
title : verbatim heading text after '## ' (trimmed).
status : the first '**Status:' line in the block, whitespace collapsed, truncated at the first sentence end, single trailing period; empty when no such marker precedes the next heading.
done_implies : the first '**Done means:' line in the block, trimmed.
record_only_ids : ids that open a record in bold-lead form (**Qnn -- ... or **Qnn (lane) -- OPEN) and have no '## Qnn' heading. A heading-only walk cannot see them, so no table row is created for one (it has no heading title and no block bounds); the manifest carries the list and the markdown table states the coverage it implies.
ids_opened_total : how many distinct ids the queue opens in any form, so rows x record_only_ids x ids_opened_total reconcile in CHECK.

Run

python3 extract_workqueue_census.py # regenerate python3 CHECK # verify against the source