09 — Objective H: Clone Lifecycle Decision

Decision: KEEP the clone as the persistent regression harness.

The clone carries the full governed-activation substrate (topic promotion, reconcile, 10-detector drift, unified scanner, issue/event boundary, UI projections) — the only environment where this layer is committed and testable end-to-end. Production cannot host it until ospa≥1. Discarding would force a rebuild.

Clone facts

DB directus_gov_test_20260602, same Postgres-16 instance, ~1202 MB. Isolated (not in query_pg allowlist; no app DSN). Contains all macro artifacts + 3 idempotent scan runs of state.

Lifecycle options & commands

• KEEP (chosen): no action.
• Snapshot: ssh contabo "docker exec postgres pg_dump -U workflow_admin -d directus_gov_test_20260602 -Fc -f /tmp/clone_topic_promo_20260603.dump".
• Reset macro layer: run G_prod_99_rollback.sql (COMMIT) then Z_rollback_A.sql on clone.
• Reset whole clone: DROP DATABASE then re-clone from prod via pg_dump|psql.
• Drop: DROP DATABASE directus_gov_test_20260602; (~1.2 GB reclaim).

Disk budget

~1.2 GB resident; compressed -Fc dump materially smaller. Acceptable; no action to keep.

Verdict

KEEP. Snapshot before next mutating macro; reset levers documented.