Curvature-Dependent CPT Framework (V1–V7): Structure, Diagnostics, and an Honest Claim Ceiling

We report an audited status of a curvature-dependent CPT-violation programme developed across versions V1–V7. The mandatory layer (V7.0 effective field theory with a white-hole pair lifetime \(τ_W\), refined in V7-D) separates fast and slow timescales and yields a three-layer adiabatic factor \(Θ\). At last scattering this removes a background pathology while leaving an infrared (IR) perturbation channel aimed at large-angle CMB anomalies. A 2026 numerical campaign tested geometric quadrupole integrals, multipole-dependent IR windows \(I_()\), Commander-native low-\(\) TT likelihoods, and a joint \(w(z)\)–low-\(\) pressure test. After correcting false positives—an incorrect Sachs–Wolfe line-of-sight distance, an over-strong flat \(I_≡ 1\), and a \(w(z)\) formula that cancelled the coupling amplitude \(ε_0\)—the honest data-level ceiling is **soft or partial lock** on non-phantom anchors. A physical hard-window \(I_()\) does **not** beat \(Λ\)CDM at a BBN-motivated bound. DESI-like \(w_0w_a\) summaries are structurally incompatible with the present \(w≥ -1\), \(w(0)=-1\) parametrization; we record that tension and do not retune \(w\) in this round. What remains is structural: timescale layering, geometric \(I_2≈ 1\) after the \(χ_*\) fix, a monotonic physical \(I_()\), a falsifiable joint-test design, and a documented diagnostic trail. What does not remain is any claim of a robust Planck low-\(\) win or a hard \(ε_0\) interlock.