LT_RC Clamping Monday Runbook

Date: 2026-04-13 (Mon, first production day after ENTRY MID_FALLBACK CLAMPED deploy)

1. Confirm the latest LT runtime log exists under Synology /volume1/aegis/logs/pt/ and is updating.
2. Run python3 aegis_v3/scripts/monitor_ltrc_clamping_live.py --once from the repo root.
3. Expected snapshot: non-zero Clamp events is acceptable; Last events should show timestamp, symbol, desired, effective, and credit.
4. Interpret ratios: desired=0.8500 is the configured discount, effective>desired means clamping lifted the limit just enough to clear the credit floor.
5. A small delta such as +0.0037 is normal; a large delta such as +0.1222 means the spread was close to the floor and needed stronger lift.
6. Start live monitoring during market hours with python3 aegis_v3/scripts/monitor_ltrc_clamping_live.py --follow.
7. Watch for recurring symbols in the last 5 events. Repeated clamping on the same name is a candidate for later price-drift review, not an immediate incident by itself.
8. After 10 or more LT_RC entry attempts have accumulated, run python3 aegis_v3/scripts/alert_ltrc_zero_entries.py --log-dir /volume1/aegis/logs/pt --date 2026-04-13.
9. Expected JSON on a healthy day: status="ok" with attempts>0 and submits>0.
10. Incident condition: status="alert" with attempts>=10 and submits=0.
11. If the alert fires once, re-run it 5 minutes later to exclude a short-lived open-delay artifact.
12. If the second run still shows submits=0, capture the JSON and the last 5 clamp events from the live monitor.
13. Generate the human-readable summary with python3 aegis_v3/scripts/ltrc_clamping_daily_report.py --log-dir /volume1/aegis/logs/pt --date 2026-04-13 --out /tmp/ltrc_clamping_2026-04-13.md.
14. Expected daily report fields: total scans, total clamping events, per-symbol average desired/effective ratios, submit count, fill count, reject count.
15. Healthy reference from the 2026-04-10 post-fix session: 59 mid-fallback attempts, 41 clamped events, 40 submits.
16. If clamp events are present but submits stay at 0, treat it as a regression of the pre-fix dead-letter state.
17. First escalation target: inspect whether rejects are dominated by ENTRY SUBMIT REJECTED or [OrderManager] LT_RC REJECTED.
18. Second escalation target: compare recurring symbols against BROKER_PRICE_DRIFT observations from 2026-04-10 and 2026-04-12 verification notes.
19. If submits=0 persists for 15 minutes after the first 10+ attempts, record the alert JSON and daily report path in AEGIS/WORK_LOG.md.
20. Escalate to the LT runtime lane as a P1 regression if submits=0 persists for 30 minutes or the report shows clamp events without any OPENED: fills.
21. Do not SSH-edit Synology files, restart containers manually, or run docker commands during this workflow.
22. Allowed actions are read-only log inspection, local script execution, git-based fixes, and CI/CD redeploys if a code rollback or patch becomes necessary.