06 — DB vs Host Drift Detector (v_rp_db_host_source_drift_detector)

Six live comparisons; classified OK / STALE_SNAPSHOT / DB_ONLY / FILE_NO_REGISTRY / REGISTRY_NO_FILE / GRAIN_DIFF / NEEDS_OWNER_RECONCILE.

drift_check	db_side	host_side	delta	class	actionable	disposition
dot_cron_declared_vs_host_crontab_mapped	42	7	35	REGISTRY_NO_FILE	yes	NEEDS_OWNER_RECONCILE
fs_dot_bin_files_no_registry	101	0	101	FILE_NO_REGISTRY	yes	T1_CAN_TRIAGE
trigger_registry_snapshot_vs_live_pg_trigger	107	408	301	STALE_SNAPSHOT	yes	NEEDS_OWNER_RECONCILE
scanner_provenance_age_vs_real_run_age	0	0	0	STALE_SNAPSHOT	yes	T1_CAN_WIRE
pg_trigger_census_vs_live_adapter	408	410	2	STALE_SNAPSHOT_MINOR	no	OK_DAILY_REFRESH
job_queue_kinds_vs_rows	8	13	5	GRAIN_DIFF	no	OK_BY_DESIGN

Notes

• 35 cron drift: dot_tools declares 42 cron-type DOTs (35 with a cron_schedule) but only 7 map to a live host crontab entry → 35 declared-but-not-on-host. Owner must confirm each is scheduled elsewhere or strip the stale schedule.
• 101 fs-no-registry: bin files with no dot_tools row; T1 can triage helper/noise vs real DOT.
• trigger_registry 107 vs live 408: the registry snapshot is badly stale; refresh from pg_trigger.
• Anti-false-positive: job_queue 8 (kinds) vs 13 (rows) is a GRAIN difference, not drift — labelled OK_BY_DESIGN. pg_trigger +2 is daily-refresh noise. Marking these non-actionable prevents crying wolf.

4 actionable drifts, 2 by-design.