What the results field coverage actually is

From when a RESULT exists at all, per country and category, MEASURED from the store rather than typed (2026-10-17): country, category, from (the date of the earliest stored result for that lane, the race's OWN date — so a Great Britain, Ireland or South Africa lane is named by its own card, not by the Sydney day it fell on; from_day and to_day above are Sydney days, which is a different question) and results (how many it holds). This is the figure to read before settling or scoring a lane. It is NOT coverage_from, which is one global value — the earliest result anywhere, currently an Australian card — and so says nothing about when a newer lane began. It is NOT field_floors either: that is when the FULL FIELD began, which can be later than the result. Australian thoroughbreds have results from 2026-08-15 and fields only from 2026-09-01, so neither value can be inferred from the other. A lane absent from this list has no result in the store at any date: races in it are priced and archived but cannot be settled here, which is true of every country outside Australia, New Zealand, Great Britain, Ireland and South Africa.

Machine-readable twin of field_scope (2026-10-15): one row per country and category that has a full-field collector — country, category, field_source (the machine value field_source carries on /v1/racing/results), source_name and from (the earliest Sydney card a stored field exists for, measured). Codes outside this list hold placegetters only.

One row per country + category over the window: {country, category, field_source, races, complete, pending, overdue, unavailable, unsupported, not_applicable}, which sum to races, plus four keys added 2026-10-19 that are NOT part of that sum: result_final (races whose result is final — placings, and dividends where the book publishes them, are in; this is independent of the full field), overdue_source_refused (the part of overdue whose source is currently refusing our collector, see field_sources_refused), oldest_overdue_age_h (hours since the start of the oldest overdue race, null when none) and refused_since (UTC, set while this row's source refuses us). Since 2026-10-20 each row also carries result_completeness {provisional, final_placings, complete_full_field, not_run}, a separate partition that also sums to the row's races.

One row per Sydney day + country + category, newest first, with the same counts. This is the day-by-day answer to 'is yesterday finished yet'.

Venue-days still missing a field in a code we DO collect: {day, venue, venue_id, category, country, settled} plus every counter above (races, complete, pending, overdue, unavailable, unsupported, not_applicable), which sum to races exactly. Ordered worst first: most unavailable, then the biggest outstanding shortfall, then oldest day. settled is true when nothing on that venue-day can still fill (no pending, no overdue) — that is the difference between a genuine terminal gap and a card simply still in progress, and without it a day mid-fill outranks a real one. A venue-day is listed when complete + not_applicable < races: an abandoned card has nothing to collect and is NOT a shortfall. A venue-day whose only gap is unsupported is deliberately absent — it would be most of the list and it is already counted above. Capped at 200 entries; venues_incomplete_truncated says when the cap bit.

Full-field sources currently refusing our collector (2026-10-19): one row per source with field_source, source_name, http_status, refused_since (the first refused run of the current streak, UTC) and checked_at (the latest run). Races of that source whose fill fell inside the refusal stay overdue and then unavailable and say so in field_note; their placings and deductions are unaffected. Empty when every source is serving.