Pre-registered audit protocols: reinforcement learning vs classical control in data-centre cooling, and physics-consistency of learned thermal surrogates

Pre-registered audit protocols (frozen before results) This deposit timestamps two frozen study protocols. Each was frozen in a local git repository BEFORE any full-matrix result was produced; amendments made before results are included in the files themselves. 1. PROTOCOL-dc-cooling-audit-v1.0.md "Does Reinforcement Learning Beat Tuned Classical Control in Data-Centre Cooling?" Frozen 2026-07-24, git tag protocol-freeze-v1.0, commit f93073f. 2. PROTOCOL-dc-thermal-physics-audit-v1.0.md Physics-consistency audit of learned data-centre thermal surrogates. Frozen 2026-08-04, git tag protocol-freeze-i4-v1.0, commit af9fc76. Purpose of this deposit: independent third-party timestamp of the pre-registration of both studies. Results, code, and manuscripts are not part of this deposit. The corresponding findings papers will cite this record's DOI.