Every page currently published across the Sanjow brand domains, read straight from their sitemaps — so QA can tell a broken URL apart from one that was never live.
ex17 finds ex-17.Sitemaps are the only source that answers “is this page live to the public”. Ads data cannot: a page with zero impressions does not appear in it at all, so an Ads-driven list would quietly omit exactly the pages most likely to be broken.
Top Up Mobile and FlyRecharge publish sitemaps past the 2 MB ceiling our tooling caches at, so they are pulled in full on every refresh. Google accepts up to 50 MB, so this is our constraint, not a search problem.
One of these hosts sits behind a CDN that serves a stale copy and ignores Cache-Control: no-cache. The console rolls a query parameter once per refresh window to defeat that, because under-reporting is the one failure it must not have — it would mean telling QA a page is not live when it is.