The Exact $O_1$ Kernel of Claude's Proposition 5.6: The Near-Diagonal Wedge, the Selberg-Integral Barrier, and an Audit of Unconditional Results

Performing exact unordered-pair symmetrization on Claude's original off-diagonal formula (conjugate-pairing algebra, not using the prime-pair conjecture, checked for floating-point consistency against 1000 randomized data sets), this separates out the leading term $O_{1,g}$ from the remainders $O_{1,\mathrm{tail}}$ and $O_{1,H}$ that must be proved separately. Setting $n=m+h$, the near-diagonal limit yields a signed oscillatory universal kernel: $\kappa(u)=(\sin2u-\sin u)/u$. The key finding is that $P_{70}$'s arithmetic input is not a single shift $h\sim X/T$ but an entire wedge: $1\lesssim h\lesssim T^{\sigma-1}$, $m\in[T,T^\sigma]$ — and under the flat taper, at $\sigma_{70}\approx1.042628$, the weight falling in $m\ge T$ is only about 0.488%, echoing the 0.114% Fourier-mass figure from the previous round. The document then personally audits two genuine unconditional results from the literature: Zaccagnini's Selberg integral unconditionally reaches $H\ge X^{1/6-o(1)}$, but the $P_{70}$ wedge's shift scale is only $X^{0.040885}$ — the ranges themselves don't even intersect ($0.040885<1/6$) — and moreover that result controls the coarse quantity $o(XH^2)$, not the signed weighted second-correlation constant that WPPH needs; the 2024 Matomäki–Radziwiłł–Shao–Tao–Teräväinen higher-uniformity result, while powerful, has an applicable range $H\ge X^{1/3+\varepsilon}$ that is still far larger than the $0.040885$ needed, and the statistic it controls is also not the weighted prime-pair second-trace asymptotic that Claude's $O_1$ requires. Conclusion: the results found so far are not sufficient to prove $P_{70}$ unconditionally through Claude's $O_1$ pipeline, because the range, the statistic, and the constant/sign all have to match simultaneously — you cannot just see “short-averaged Hardy–Littlewood” and declare it already established.