Temperature-constrained pulse-power limits in parallel microchannel cooling with unresolved flow maldistribution

Integrated numerical code, retained results, and manuscript sources supporting the study of temperature-constrained power ceilings in parallel microchannel arrays with unresolved flow maldistribution. Exact-version DOI: 10.5281/zenodo.23057309. This release consolidates the earlier reproducibility materials and the September 29-30, 2026 additions into integrated_reproducibility_20260930.zip. To support unreliable upload and download connections, the ZIP is distributed as numbered raw byte parts of at most 50 MB each. Download all parts, UPLOAD_MANIFEST.json and one of the supplied reassembly scripts into the same folder, then follow README_REASSEMBLE.md. Each part and the restored archive are checked with SHA-256. The parts are not independently extractable ZIP files. This transport arrangement removes no scientific data and restores the complete original archive. The restored archive includes reproduction instructions, environment requirements, verification records, and file-integrity manifests. Earlier published releases remain accessible through the version history. The added evidence includes expanded eight-channel allocation sampling, critical uniform-edge pulse sampling, independent full-order linear checks, nonlinear periodic reduced-order basis-enrichment checks, and eight full-temperature-state nonlinear periodic cases with time-step refinement and independently evaluated power roots. Matched comparisons quantify the power-ceiling bias associated with different allocation-selection assumptions. Updated figures, manuscript sources, and supplementary documentation are included. The reported power ceilings, derating values, and ordering changes are conditional finite-sample results for the specified models, observations, heat sources, and pulse schedules. The nonlinear full-order calculations retain a quasi-developed thermoviscous hydraulic closure; they do not solve the full momentum equations, certify a continuous global worst allocation, or validate device ratings. No new experiments were performed. Comparisons using digitized published data are attributed in the documentation. Third-party full-text articles are not redistributed.