Internal audit and limits

• Checked the signed coercivity argument on the compact unit slice, including an empty zero-quadratic locus and the zero cone.
• Retained the affine-input Dickson argument and the nonnegative mixed term; did not infer all-input convergence solely from unit convergence.
• Checked necessity with rational scaling into the actual auxiliary lattice, not saturation of the averaging lattice.
• Separated growth at infinity from positivity at every nonzero lattice point. Signed angles do not give constant coefficient one.
• Replaced the unit detector by a nonzero homogeneous monomial value in the diagonal grading argument. Both relation sides remain intact.
• Corrected an intermediate exponent in the double-sum derivation: it is j(j-1)/2-bj+k, not j(j+1)/2-bj+k. The former agrees with the Euler numerator (q^-b;q)_infinity and the exact coefficient test.
• The abstract one-block control has theta=(2,-1,0) and arbitrary algebraic moment character; no actual manifold incidence is claimed.
• The accompanying rational recurrences and double-sum checks are different computations by this same authoring agent, not independent-model or external review.
• No claim that ray positivity implies nonzero F on every actual triangulation, no unrestricted conjecture resolution, no new full Publication.

Next: nonzero native monomial detection under the signed criterion, preserving the actual image lattice and signs. The formal proof and independent-reconstruction maturity gates remain unmet.