Question

You run logical replication from your primary Postgres to a separate analytics/CDC target. Suddenly the subscriber falls hours behind: the downstream warehouse is stale, lag on the slot grows monotonically, and WAL is starting to accumulate on the primary (raising disk concern). The primary's own OLTP traffic looks normal. Earlier today a maintenance job ran a single `UPDATE` that touched ~80M rows in one transaction on a heavily-replicated table. Triage and mitigate.