Lineage-aware stale-table alerting (suppress downstream noise)

When an upstream source stalls, every table that depends on it also goes stale, producing a storm of alerts. This workflow reads your declared lineage, identifies which stale tables are root causes versus downstream victims, and alerts only on the roots, listing the affected downstream tables as context.

Use it when your warehouse has deep dependency chains and on-call gets buried in correlated freshness alerts during a single source outage. It turns 30 alerts into one actionable root-cause page.

1. 1A schedule triggers the freshness sweep.
2. 2A BigQuery query returns the last-refresh time for all monitored tables.
3. 3A query against your lineage table loads parent/child edges between those tables.
4. 4A logic step marks each stale table, then walks edges to keep only stale tables whose parents are fresh (the true roots) and groups their stale descendants.
5. 5A Slack message posts each root cause with its blast-radius list.
6. 6A PagerDuty incident opens only if a root sits in the tier-1 critical set.