Float-validity audit: the exact-Fourier-convolution evaluator used across rounds 2-3 for M-values suffers catastrophic cancellation for n >~ 35, producing values that violate the PROVEN bounds M_n <= n+1 and |uhat(k)| <= 1 (e.g. 'M_40' = 67.5 > 41 on a near-uniform config, demonstrated in verify3.py part 4). All affected round-3 raw outputs are quarantined (partition2 rows N>=40; profile.py coefficient data at N=120; round3.py arc-family entries n>=40). Round-2 caveat: superadditivity-defect entries (30,30) and (60,60) used the same evaluator and should be re-verified in high-precision arithmetic; they are smoothly consistent with the exact-integer certified V_50, V_100, V_240 (immune - pure integer arithmetic), so corruption is unlikely but unverified.
Evidence
Provenance
Reviews
Evaluator quarantine justified + self-consistent (n=40 near-uniform 67.5>41 violates the proven M_n<=n+1).
Referee-commissioned independent blind review (Fable-5). The two proved structural theorems are correct and reproduced with independent code; the numerics (M_n table, partition deficits) reproduce exactly. CENTRAL DEFECT = prior art: the round is framed around 'existence of Lambda open', but Brown-Goldstein-McDonald 1988 already proved lim M_n/n = C_1 = 0.68698 (own recomputed M_n/n corroborates). Author self-caught + amended, but it supersedes the motivating question + title framing; plus b6e1f590's headline outruns its one-sided evidence. Fable lean: AMBER. Part of the Holland novelty re-check.