Field Audit: Board-Complete Data and the Limits of External Gravity-Model Testing

The Hermes gravity relation predicts galaxy rotation curves from observed baryonic structure and a galaxy-age term with no per-galaxy free parameters. Previous work evaluated Hermes on a 133-galaxy age-verified subset of the SPARC database. This paper documents the attempt to test Hermes on independent, non-SPARC data. The attempt exposed a data standardization problem. Fair cross-model rotation-curve testing requires board-complete inputs: observed velocities, uncertainties, and independent per-radius gas, stellar disk, and bulge components with documented mass-to-light ratio conventions. We audited public databases, survey products, author files, and doctoral theses. Most candidate paths failed as model boards before any model could be run, each for a different reason: calibration-set overlap, incompatible velocity targets, missing baryonic components, privately held data, and convention mismatches. The BIG-SPARC project independently identifies this limitation but is not yet publicly available. One external board was reconstructed: M33, from the published baryonic mass model of Corbelli et al. (2014). On this Format B board, Hermes produced a non-catastrophic but elevated fit (χ²ν ≈ 3.2), while Modified Newtonian Dynamics fit cleanly (χ²ν ≈ 0.1). The Hermes residual concentrated in the outer disk, a region independently documented as dynamically complex. Baryonic-convention and age diagnostics narrowed but did not eliminate the gap. The result is one external board, not a general verdict. Its value lies in three findings: the public data ecosystem is less board-complete than commonly assumed, Hermes failures can localize to physically meaningful boundaries, and rigid models can expose structured input problems that flexible models may absorb. External tests of any baryonic gravity model will remain as much a test of the data infrastructure as of the model itself until standardized board-complete datasets become available.