Per-operator timetable freshness and scrape schedule (backs /status).

One row per operator (GTFS agency) in the graph data: stops, routes, trips, trips running in the next 7 days and in the planning window, timetable validity (service_start / service_end, days_left) and a status computed against today — ok | expiring (ends within expiring_days) | no_service (valid, but no trip in the next 7 days) | expired | unknown (no calendar). at_risk marks a timetable that ends before the next scheduled source check + 7 days. feeds carries provenance per feed (source_url, licence, when the file content last changed, quarantine reason); units the scrape cadence, last check, its result and the next check; graph the last planner rebuild and the earliest next one. On a row that is not ok (or is at_risk), action says what the daily freshness watchdog is doing about it — refreshing | update_pending (newer timetable fetched, live after the next rebuild) | superseded (another feed for the same network replaces this copy) | scheduled | no_new_timetable (publisher has released nothing newer) | checked (reviewed, e.g. seasonal) | removing | needs_review — and watch summarises the last check.