# Audit v3cf8 (auditor, 25 Oct 2026)
Verdict: NOT VERIFIED (FIX/wait, KILL-lean).
- Clean fwd (entries after look-ahead fix 25 Oct 01:20 UTC): n8, mean 1.17x (below the n>=15 gate; too few to trust).
- All fwd n140 mean 1.01x (54 zero exits, biased pre-fix); replay n503 mean 0.95x. Replay under 1.0 and fwd barely above: no demonstrated edge. Siblings v3cf6/7/9 same family killed at 0.74-0.99x.
- Sellability / outside sellers / hand rebuild not run: pointless until clean mean >= 1.1 at n>=15.
Re-audit at clean fwd n>=15 with mean >= 1.1 and lower80 >= 1.0; otherwise testmgr retires at decide-by 28 Oct.
# Audit v3cf9 (auditor, 25 Oct 2026)
Verdict: NOT VERIFIED, KILL (mean far under 1.0).
- Clean fwd (entries after look-ahead fix 25 Oct 01:20 UTC): n3, mean 0.47x.
- All fwd n140 mean 0.74x (83 zero exits); replay n520 mean 0.73x. Both far under 1.0 and agree, so no edge; wider age/stop did not help vs siblings v3cf6 (killed 0.99x) and v3cf7 (0.93x).
- Sellability / outside sellers / hand rebuild not run: pointless while mean < 1.
Testmgr: retire v3cf9 now (kill rule: mean<1 with upper80<1.05 already true at n140).

## Re-audit 25 Oct 22:00 UTC
Verdict: KILL. Clean fwd n18 (entries after 25 Oct 01:20 UTC) mean 0.87x, 3 zero exits; gate was n>=15 mean>=1.1. Replay 0.95x, siblings v3cf6/7/9/12 all killed. Testmgr: retire.
