A Gaia halo-tracer × Fermi γ-ray halo-residual spatial cross-correlation: why a rank cross-correlation with a pure spherical monotone template cannot provide dark-matter-specific evidence.

Version 7.0 supersedes version 6.0 after a fifteen-run adversarial epistemic audit (~560 audit rounds, 119 edit waves, two consecutive five-round certificates without a critical or medium finding) and an independent external scientific audit (8 September 2026, "major revision"). It is published as a methods paper. Evidential content: (1) the structural result — for a pure spherical monotone halo template, a rank cross-correlation measures central concentration and cannot provide dark-matter-specific rank-morphological evidence (an NFW² J-factor map has Spearman −1.000000 with the Galactocentric angle); (2) the corrected-support measurements — on the coordinate-defined 7,596-pixel box the latitude-controlled partial rank correlation with the Totani (2025) halo residual is +0.182 (p = 0.014, type-I error measured), falling to +0.016 (p = 0.44, one-sided exceedance, the single convention of this version) once the Galactocentric angle is controlled; the primary tracer retains 8.8% of its link, the other tracers ×1.1–1.8 more on a like-for-like block. The outcome-conditioned 5,808-pixel support of earlier versions is quarantined to history. Exploratory and conditional (stated as such throughout): the characterisation of the residual structure S at 20 GeV in an independent pipeline's residual; the decomposition of the two-pipeline correlation into a shared-photon and a structural part (mechanism supported, size unresolved); the dark-matter channel statements of the loopholes appendix (μμ closed at 500–1000 GeV; ττ at 1 TeV and both neutrino cells UNSETTLED under their full budgets). No detection is claimed. Changes since 6.0: title and central claim narrowed; single p-convention; 47% shared-photon share removed from the abstract; grades (confirmed / supported / unresolved) for every nature-of-S statement; data/code/competing-interest statements; revision archaeology moved to a machine-readable CHANGELOG (256 entries); a locked environment specification, input and script manifests with SHA-256, and the provenance of the PPPC4DMID neutrino/positron tables added to the capsule; the two audit certificates with the full finding ledger (6,767 entries) included. A continuation paper (forward-modelled Gaia selection with mocks, structured surrogate nulls, model-error ensemble, signal injection, a single multiplicity family, empirical confidence intervals, a photon-disjoint cross-fit, containerised runner) is in preparation; the Totani excess itself has since been reproduced at pixel level (arXiv:2607.08552) and its dark-matter interpretation is in tension with AMS-02 antiprotons (Wang & Duan 2026).