Newest records first. Older freezes stay.

Massless Hartle–Hawking scalar, M=ℏ=1, minimal coupling for the tensor comparison. φ²=1.9571596066001755×10⁻⁴. Box(φ²)=3.1414970956971594×10⁻⁷. Scalar-inferred trace −1.2088044476173205×10⁻⁷. Stored PMR trace −1.2086496991741827×10⁻⁷. Residual 1.5474844313788875×10⁻¹¹.

Scalar-only allowance 5.251546195221486×10⁻¹² does not cover the residual. Inherited PMR allowance 3.856578844969775×10⁻⁹ dominates the combined allowance 3.861830391164996×10⁻⁹, which does. Empirical, retrospective, target-exposed. Not a confidence interval or a tail bound.

Primary: N=24, L=48, Duffy 256. Rectangular grid excluded. Nine radii 5.92–6.08M support the derivative only. Two corrections: omitted massless log and rational tails, and the archived static boundary coefficient. This website did not rerun the calculation.

A separate derivation matches a, b, and e exactly, with c=d=0. Sixteen checks recorded as passing. Target-exposed, not blind. Not a second Christensen/AHS implementation.

Replay of derived e on five stored radii is recorded as unchanged to roundoff. r=6 bitwise equal to the stored tensor in that report. Stored samples are not independently validated components.

Stored-tensor Box prediction 3.141187598810884×10⁻⁷ is distinct from the 9 October independent Box.

Normalized multi-frequency self-cancellation was already implemented. Component values at r=5.5, 6, 6.5, and 7 agree 12/12 with digitized published curves within stated tolerances. Those table values are not in the public JSON.

Full chain remains MRX_R_INVALID. r=5 under analysis v1.0 remains ANALYSIS_R_INVALID. Supplementary v1.1 A11_C2P_PASS_ALL does not replace it.

Stored r=6 conservation residual 8.23579665749×10⁻¹⁰ versus primary tolerance 2.08667476468×10⁻⁸. Primary-tolerance pass of the stored diagnostic. WARD stays HOLD.

Unseen window ω∈[10.40,10.90], 51/51 angular-accepted, ε₁=37.5 fixed, never fitted. Accepted record reports C1–C6 PASS. C5 corrected from sequential residual fitting to joint-H1 source A/B transfer plus a fresh target cubic.

Accepted directional R² 0.9988049064965246 and 0.9985764043986991. 9 October replay 0.9988049064965255 and 0.9985764043986983. Roundoff only.

Not global asymptotics, not an infinite-frequency integral, not full HH RSET. Discovery D=FAIL and F=CRITERIA_GAP remain. Grid file is not published. C1–C4 and C6 definitions are not in the public package.

51/51 ANGULAR_ACCEPTED at r=6M, ω∈[9.8,10.3], Δω=0.01. ε_n={37.50,70.17,102.8,135.5} declared before examining the scan and never fitted. Fixed ε₁=37.5 H1 explains R²≈0.997 of F_reg^tt versus ≈0.084 for a smooth cubic. Dominant residual projection is ε₁ by 6.72×. Split-window Δφ≈0.035 rad. tt/rr/θθ all R²≈0.997 with θθ ≈π-shifted.

Old surrogate discriminator: INVALIDATED_DO_NOT_USE_FOR_CNG_DECISION. Offline re-score: A/B/C/E PASS, D FAIL retained, F=F_CRITERIA_GAP. Overall OFFLINE_RESCORE_UNCLASSIFIED_CRITERIA_GAP.

Say substantial numerical evidence for the predicted ε₁ CNG oscillatory structure, as of this date. The window that was next was computed the following day and received a narrow pass. That later pass does not rewrite D=FAIL or F=CRITERIA_GAP.

Analytic Levi–Ori t-splitting mapping was frozen before reading the ω=8,10,12 |F_reg| ratios as a derivation. F_sing removes the local coincidence polynomial (ω³ and ω on Schwarzschild). Remote null-geodesic / caustic singularities remain as oscillations of F_reg. Levi (2017) at r=6M states that F_rr^Reg is dominated by oscillations with amplitude ∝ ω^{5/2}. The ordinary integral need not converge; the generalized integral is the method.

Referee verdict: GROWTH_EXPECTED_PRE_GI. DECAY_EXPECTED_POST_SUBTRACTION is not the literature prediction. A missing leading ω³ term is disfavored by the certified ω=8 cancellation |F_reg|/|F|_tt ≈ 0.0124. Do not fit a power law on {8,10,12}.

Do not run Abel. Do not automatically extend to ω>12. Not HH RSET. Not frequency-support convergence.

Both ω=10 and ω=12 received ANGULAR_ACCEPTED and confirmation under the frozen adaptive angular-convergence gate with predeclared L_ceiling=180. Certified pointwise domain to ω=12 is PASS, conditional on the ℓ gate.

|F_reg^tt(10)|/|F_reg^tt(8)|=2.18019. |F_reg^tt(12)|/|F_reg^tt(8)|=4.02997. DECAYING_HIGH_OMEGA_TAIL_DEMONSTRATED=FALSE. ASYMPTOTIC_REGIME_IDENTIFIED=FALSE. FREQUENCY_SUPPORT_CONVERGED=FALSE.

This is not a dense spectrum, not Abel, and not an HH RSET reproduction. L* for 10 and 12 is not published in this freeze.

The reusable adaptive angular-support gate was tested on previously unused ω=7, starting deliberately under-supported at L=30. It selected L*=90 without a hand-chosen final ℓ_max. Confirmation at L=105 left F_reg unchanged within the frozen criterion.

Important because the gate was not merely fitted retrospectively to already-inspected ω={4,5,6,8} points. ANGULAR_GATE_VALIDATED_PROSPECTIVELY = TRUE.

An ANGULAR_ACCEPTED point is merely eligible to enter a future controlled frequency-integration dataset. Not Abel. Not HH RSET.

Acceptance rule frozen before the run: all of tt / rr / θθ must satisfy D_ℓ < 0.05 for two consecutive ΔL=15 terminal windows while cancellation remains healthy.

L=75 vs 60 FAIL. L=90 vs 75 PASS (D_ℓ ∼ 10⁻¹²). L=105 vs 90 PASS (D_ℓ = 0). The ω=8 issue was late angular settling of the same insufficient-ℓ-support problem, not a separate systematic.

A controlled isolation changed ONLY ℓ_max at fixed r=6M, ξ=5, ω=8, Hartle–Hawking, same Levi–Ori F_sing, same equations / BCs / precision.

|F_reg|/|F|_tt: 8.698 at L=25, 3.775 at L=35, 1.011 at L=50, 0.0124 at L=75 with F/F_sing_tt = 1.013. Cancellation healthy in tt, rr, and θθ at L=75.

This is not proof that the full PMR RSET is correct. Historical high-ω FAIL is retained as the first chronology step.

Earlier sparse high-frequency probes using approximately fixed ℓ_max ≈ 35 appeared pathological for ω ≳ 5: regularized residuals grew, F/F_sing deteriorated, and |F_reg|/|F| became O(1) and eventually >1.

A precision/rtol ladder showed the failure was reproducible and was not ordinary float64/ODE noise. Later isolation as insufficient ℓ-support does not erase this FAIL.

Old under-resolved high-ω points must not be used as production Abel input.

A third internal mode-sign derivation brought INTERNAL_SIGN_DERIVATIONS = 3/3 PLUS. Public archaeology of the authors' private numerical implementation remains UNKNOWN. SIGN_VERDICT = UNRESOLVED.

The mixed-time sign lane is no longer the live B1 frontier. It remains externally reviewable, not confirmed, paused. Production sign is not flipped. Author note remains UNSENT.

Two fresh implementations — neither importing AG RSET / production modules — independently recover the analytically DERIVED overall PLUS mode contribution under the printed source conventions. Independence grep: PASS. 5/5 synthetic + 18/18 frozen-real = 23/23 DERIVED; 0 PRINTED; 0 unresolved.

MODE SIGN is independently reproduced. LOCAL/GEOMETRIC BRACE is a source-analytic audit, not an independent reconstruction of every geometric brace.

A displayed-formula issue does not imply that the authors' published numerical RSET values are wrong.

The independent Mτ materials were reduced to a public minimal pack: 17 inner files, 0 PDFs, no AG runtime imports, no emails, no IPs, no credentials. SHA256 of the published tarball is recorded on the study page.

Author-contact note remains DRAFT / UNSENT. Authors have not been contacted. This website freeze does not send mail.

A sign tension was identified between the displayed mixed-time coincidence formula (TBO Eq. 18 / analogous ABOT Eq. 16, overall MINUS) and the sign obtained directly from the papers' own Euclidean Green-function definition, Δτ convention, W = G − K relation, and Euclidean metric (overall PLUS under this reading).

This was not accepted on the basis of Astral Gate's production code. Phase D published numerics remain UNKNOWN. Archaeology of private code remains UNKNOWN.

A candidate explanation that the project's point-split referee operator itself carried a wrong second-derivative sign was tested and REFUTED. The operator evaluates MINUS_D2. Finite-Δτ referee: Mτ_PS = − Mτ_printed at all four radii, consistent with the displayed-formula tension, not with an internal operator bug.

Nested radial-spacing refinement passes for both first and second derivatives at all four certification radii ξ = 1.2, 1.8, 3.0, 5.0. Historical nested-h HOLD is retained as dated history, not deleted.

At ξ=1.2 the coarse D(h) value is unavailable. Certification there is h/2 → h/4 stability, source-derived modewise agreement, N/L spot referee, and mixed error budget — not a measured three-level Richardson order.

Closing this derivative gate did not promote RSET. October 9: this PASS is the old implementation and tolerance, not the repaired scalar budget.

Radial Green, zero mode, high-mode conditioning, complete m=6 Hadamard subtraction, ⟨φ²⟩ convergence, source-derived radial derivatives, same-ladder derivative comparison, and nested-h refinement have passed at ξ = 1.2, 1.8, 3.0, 5.0 under the implementation and tolerances of this date.

The renormalized stress tensor itself has not been assembled. B1 reproduction remains unclaimed. DIRECT_LOG_5E3 FAIL_EDGE is preserved beside DIRECT_LOG_1E2 PASS. October 9 reassessment: do not read “complete m=6”, “no blockers”, or a blanket TRACE PASS as a current universal claim.