MART/KTA RMEM2 Minimal Relativistic Closure: Audit-Level Reproducibility Pack for Planck low-l, Lensing, BAO, SN, and RSD Tests

This release is an audit-level reproducibility package for the MART/KTA RMEM2 minimal relativistic closure. The package contains the RMEM2 closure documentation, audit records, likelihood scripts, Cobaya configuration files, external mu/Sigma tables, MCMC pilot chains, current package manifests, internal and external SHA256 integrity references, and legacy-isolated diagnostic material. The final RMEM2 closure used for the reported results is: Sigma_KTA(k,z) = 1 + Delta_KTA(k,z)mu_KTA(k,z) = 1 - R_mem(k,z)^2 * (Sigma_KTA(k,z) - 1)eta_KTA(k,z) = 2 Sigma_KTA(k,z) / mu_KTA(k,z) - 1 This replaces the older exploratory growth-boost relation mu_KTA = 1 + R_mem. Older diagnostic scripts and outputs containing this obsolete relation have been moved into legacy_diagnostics/ and are retained only for development transparency. They are not part of the reported RMEM2 reproduction claim. The primary reproducible joint comparison uses Planck low-l TT/EE, DESI BAO, Pantheon+ SN, and RSD f_sigma8. The RMEM2 closure yields chi2 = 1893.9000 compared with LCDM chi2 = 1892.2825, corresponding to Delta chi2 = +1.6175 and Delta AIC = -4.3825 for k_MART = 3 versus k_LCDM = 6. Using the same effective parameter-count convention, k_MART = 3 versus k_LCDM = 6, and an approximate joint data count N ≈ 1733, this corresponds to Delta BIC ≈ -20.75. The BIC value is reported as an approximate information-criterion diagnostic because the effective data count of the Planck low-l likelihood is not uniquely represented by a simple scalar N. A lensing-extended joint comparison is also included in the package. It yields chi2 = 1902.4713 for RMEM2 compared with chi2 = 1904.0372 for LCDM, corresponding to Delta chi2 = -1.5659. The included MCMC chains are audit-level pilot chains with finite max_samples settings. They document current best-fit minima and workflow consistency, but are not claimed as fully converged final posterior constraints. Planck high-l validation is not claimed in this release. High-l test files, if present, are retained only as transparency and infrastructure records and may contain chi2 = inf. The package includes the RMEM2 external mu/Sigma table and YAML activation metadata. A complete third-party patched CLASS/Boltzmann-solver rebuild proof, including exact solver source state, build logs, environment specification, and table on/off smoke tests, is deferred to a future solver-reproduction release.