Public endpoint - data availability stats.

Cached 5 minutes. The underlying query takes ~10s (DISTINCT source

• MIN/MAX game_date over 33M rows) and the answer changes slowly enough that every poll re-running it is wasteful. Customers surveying capabilities don't need second-resolution freshness on a "what bookmakers do you cover" answer.

Stale-while-revalidate semantics: if the cache is fresh (under 5 min), return it. If it's expired but still present, return the stale copy AND spawn a background refresh so the next request hits a fresh value. If it's never been populated (first call on a cold worker), do the full query inline - the warmup loop catches up within ~13 s of spawn so this only affects requests in that narrow window. iter_031/033 #327/#329 HIGH timeouts.