Computational supplement to "A Separating Test for Renormalization-Group Realism Using Classical Mechanical Models: Preregistered Criteria and First Results"

Code, computed results, figures, reports, a verification notebook and the datedpreregistration record behind the preprint of the same name. A Newtonian chain-and-oscillator model is shown to reproduce the full apparatus ofrenormalization-group universality — a fixed point, a universality class,insensitivity of the infrared to the ultraviolet, a finite budget of relevantparameters — with a Gaussian fixed point (outcome C of a three-way fork fixedbefore the computation was run). Two further preregistered blocks are included: anFPUT/KPZ block, which reaches nontrivial universality in the same Newtoniansetting, and a decomposition-relativity block, which shows the universal constantand the universality class to be relative to the system/bath cut. A single entry point, run_all.py, recomputes the core and decomposition blocks inabout eighty seconds and then re-checks forty-one headline numbers against thevalues reported in the paper, printing a pass/fail table. In a fresh virtualenvironment those blocks reproduce bit-for-bit. 240 built-in analytic-versus-numericself-checks run on every execution. A Jupyter notebook (notebooks/) recomputes thecore from first principles in one self-contained document, loading nothing from thedeposited results, and prints 33 registered checks. MANIFEST.sha256 carries the checksum of every deposited file. Two limitations are stated plainly rather than smoothed over. The rawstructure-factor arrays of the FPUT block (five files, 481 MB) are notredistributed; the exact commands and wall times that recreate them are given.Third-party copyrighted source texts are not redistributed; all sources are openlyavailable and listed with their arXiv and PhilSci-Archive identifiers. The singleexception is a numerical table of the Prähofer–Spohn scaling function, reproducedwith attribution and expressly outside this deposit's licences. Licence: MIT for the code, CC BY 4.0 for data, figures and reports; see LICENSE.md.