Observing Case: Does the Galactic Acceleration Scale Evolve? Deep-Regime Rotators at z ~ 2-2.6 (Version 2: corrected target ledger, separated rivals, honest odds)

VERSION 2 (2026-09-25) supersedes version 1. Version 1's candidate ledger carried values that contradict their own cited sources (largest: A68-HLS115 is at z = 1.5869 with M* = 8.1e9 Msun, not a z = 2.49 deep-regime candidate); it merged two rivals (the LambdaCDM-emergent rise was labelled 'proportional to H(z)'); and its 'one galaxy at +/-0.13 dex decides at 20:1' ignored intrinsic and halo-to-halo scatter (corrected: about 4:1; three galaxies at +/-0.10 dex or four at +/-0.20 dex are needed). Version 2 re-reads every ledger row from its primary paper, quotes the three hypotheses separately at z = 2.5 (constant 0.00; LambdaCDM-emergent +0.23 to +0.46 dex; a0 proportional to H(z) +0.58 dex), fixes a sign convention, and adds a caveat: the framework's own observable zero point is not yet sharp (a retained cold halo can raise it by about +0.2 dex; not yet computed), so this case tests the three hypotheses, not the framework as a whole. Companion to the pre-registered measurement (doi:10.5281/zenodo.22563139), written for observers. Contents: the three hypotheses at the target redshifts; the gate (window relaxed to z >= 2, lens gate tightened to 0.05 dex); what one to four galaxies decide, with the full-kernel error amplification near the gate; the verified candidates: OLAS M0717-02064 (z = 2.07, the only rotation-supported deep-regime-plausible object at z >= 2, whose published kinematics favour none of the three hypotheses), Abell 68 C4 (stellar mass disputed by 0.7 dex), SL2S J0217, and two high-acceleration controls; per-object precision about 0.43 dex from velocity data alone (gas-limited) and about 0.26 dex with a +/-0.12 dex gas mass; the decision rule. AI-assisted research programme; not peer reviewed.