Supplementary material: Good codes can be bad records -- deterministic certificates, reproduction code, and the verification trail

Complete, self-contained verification trail for the companion paper "Good codes can be bad records: when coarse-graining destroys objective redundancy." Every reported quantity is reproduced by thirteen deterministic seeded certificates. The original six-host battery (seed 20260704, at the archive root) supplies the persistence battery, the skeptic radius-2 re-check, the maximum-scalar re-run, the hub witness, and the extensive-criteria re-call. The adversarial stress battery (seed 20260705, in the tloc1_attack/ subdirectory) adds the counterexample search (including the 3D-lattice collapse), the blocking-semigroup matrix (eight schemes), the recovery-criterion matrix (six rules), the adversarial host zoo, the size ladder to N=10^5, the 42-host tree zoo, the sandwich verification, and the cap sweep (the cap-depth-dimension law). The bundle ships every generator script plus all shared machinery (host builders, coarse-graining, coordination and readability measures, stabilizer record-check predicates), so each certificate regenerates from the bundle alone (Python 3 with numpy, scipy, stim). It also includes the authoritative source/adjudication trail and a sha256 MANIFEST. Licensed CC-BY 4.0.v2.3 (2026-07-05): adds the large-volume continuation of the cap sweep (l1b_large3d: cubes to 60^3, N to 216,000, caps to w=64; pipeline validated bit-identically against the published 16^3/22^3 rows). The cap–depth law's linear form survives at every size but its coefficient is size-dependent (w≈2K+6 is the 22^3 value; the bulk-3D cost is nearer w≈3.4K); coordination growth is measured non-saturating on high-trust data (the 22^3 rollover was finite-size); and the bulk-3D retained-density crossing also pins cap-invariantly, so the robust geometric-vs-arithmetic separator is the admissible-fraction cap-law slope plus the coordination growth mode, not the extensivity crossing alone. Theorems unchanged. v2.4 (2026-07-06): folds in the SEQ-1 sequel (seed 20260706; certificates seq1_wpv_bulkslope, seq1_wpe_expander, and the adversarial re-derivation seq1_skeptic_checks) and corrects one published coefficient. The bulk cap–depth slope had not converged at 60^3: continued to 80^3 (N=512,000) and 96^3 (N=884,736) it keeps dropping, 0.297→0.236→0.214, so the robust statement is only bulk w > 3.4K (a 1/L extrapolation gives ≈0.16, bulk w≈6K as a rough midpoint, spanning w≈4K–14K across defensible ansätze); the published “w≈3.4K” was a not-yet-converged finite-size reading. The Δκ increment diagnostic stays non-saturating at both new sizes, reinforcing the depth-relative (C) reading. A matched-volume expander re-run (N to 216,000) confirms the densification-mode separators survive at scale: κ_mean growth stays geometric on the pre-completion window (a complete-graph ceiling κ_max=N−1 contaminates the full-window fit and is excised on principled grounds), and the surviving discriminators are the growth mode plus the curvature of the admissible-fraction cap law (expander sub-linear/saturating vs cubic super-linear/accelerating) — not a linear-vs-log fit split, which fails at scale. K*_ext cap-invariance at 3 is confirmed but is not an expander-unique fingerprint (bulk cubic pins there too), and the raw slope gap narrows with volume (0.355→0.220), so slope magnitude carries no weight. Outlook successors (v, volume dependence) and (vi, matched-volume expander) are thereby resolved numerically, leaving only the closed-form cap–depth derivation open. A one-line cross-reference points to a companion note on the holographic-host typing. Theorems unchanged.