DES-SN5YR Foundation Low-z Covariance-Aware Robustness Packet: Overlap-Aware Consistency Diagnostics

This dataset records a covariance-aware robustness analysis of the DES-SN5YR low-redshift Foundation subset. The analysis focuses on the Foundation-only sample with IDSURVEY = 150 and zHD < 0.060, containing 92 supernovae. The baseline Foundation-only directional residual fit gives Δχ² = 14.945456, amplitude = 0.101681 mag, RA = 20.366097°, and Dec = 57.880151°. A sequence of first-line robustness checks was performed, including survey jackknife, Foundation internal jackknife, random-sky null testing, residual sign-flip null testing, bootstrap and leave-10-out resampling, redshift-split sensitivity, local-flow / VPEC sensitivity, leave-one-object influence testing, Pantheon+ object-overlap checking, Pantheon+ Hubble-flow-overlap splitting, and HF/non-HF vector-combination diagnostics. The Foundation-only signal is internally robust across these checks. It is not explained by one high-leverage supernova, by the highest-|VPEC| objects, by a single redshift quartile, by random reassignment of sky positions, by residual sign randomization, or solely by the 51 Foundation objects overlapping the Pantheon+ Hubble-flow subset. However, the Foundation subset is not independent from Pantheon+ low-z at the object level: all 92 DES Foundation low-z objects overlap Pantheon+ low-z all, and 51 of 92 overlap the Pantheon+ Hubble-flow subset. Therefore, the DES/Pantheon+ directional agreement should not be described as independent object-level replication. The most defensible interpretation is an internally robust, overlapping-sample, cross-pipeline / covariance-aware consistency signal in low-z Foundation/Pantheon+ data. No cosmological anisotropy, preferred-axis, URSF-R3 confirmation, local-flow elimination, calibration-systematics elimination, or independent DES/Pantheon+ replication claim is made. This release is a robustness and reproducibility packet only. It is not a discovery claim or a cosmological anisotropy claim. AI assistance was used during the preparation of this robustness packet to help organize analysis checkpoints, draft documentation, summarize diagnostic results, generate plotting/checkpoint scripts, and maintain clear guardrails around interpretation. All scientific claims, numerical outputs, code execution, dataset handling, and final decisions were reviewed and controlled by the author. AI assistance did not replace independent verification; it served as a documentation, coding, and analysis-support tool.