Compact the audit log into hourly chunks in staging, verify against source row counts, and then ask the operator before cutting over to production.
ContractNo event drop. Drift threshold 0.5%. Operator approval before production write.
- 1postgres.executeSuccess
Acquire advisory lock on audit.logs
→ lock_id=8841 acquired in 0.1s
0.1s120 - 2agent.audit-writerSuccess
Sketch hourly chunk window proposal
→ Window: 60m, max_payload=4096. Approved against staging schema.
4.2s2.8k$0.04 - 3postgres.executeRunning
Backfill hourly aggregates in staging
→ 144,019 / 145,225 source rows so far
running 1m12s - 4human.operatorPending
Approve production cut-over once drift < 0.5%