Diagnostic canary. The Liveness KPI runs a real query against the database — if it shows a number, the QuickSight rendering pipeline is healthy and any blank visual on another sheet indicates a data or SQL issue. If the KPI is blank, QuickSight itself is broken.
Count of public-schema tables. Real query against the database via Direct Query -- if this shows a number, QuickSight's rendering pipeline is healthy. Blank means QuickSight itself is broken (not the data, not the SQL).
Row counts + most-recent date for the matviews + base tables this dashboard depends on directly. Per-app scope by design — Executives reads only 2 base tables; L1 reads ~12 matviews. For total deploy freshness, check every app's App Info sheet. Freshly-loaded matviews showing 0 = ETL hasn't refreshed them yet. If a base table's latest_date moves past a matview's, the matview is stale — re-run recon-gen data refresh --execute.
Most recent emitted balance day from the feed — live query against <prefix>_data_anchor (DK.1). When this lags the as_of (at emit) bullet below, the ETL has aged since the last recon-gen json apply; dashboards are still showing the deploy-time anchor, not today.