Is There a Crisis in Calibration? A Guardrailed Decomposition of Supernova Residual Structure into Selection–Calibration Transfer

I test an operational decomposition of Type Ia supernova (SN Ia) Hubble–residual anomalies into (i) expansion– history contributions and (ii) selection–calibration transfer contributions arising from survey, pipeline, and population covariates. Using a fixed sequence of adversarial diagnostics—baseline removal, SALT2 leakage regression, survey fixed effects, footprint stress tests, blocked permutations, BAO–only w0–wa projection searches with globalized p–values, and learned correction operators—I find that strong ordered residual structure and sky anisotropy can be removed by admissible correction models tied to selection and calibration features. After correction, the residual field exhibits no remaining support for attributing the anomaly to expansion–history deviations within this framework. Host mass shows negligible association with cleaned residuals, while light–curve population parameters (x1, c) retain small but statistically significant correlations consistent with selection–mediated population drift. I formalize this workflow as a guardrailed admissibility criterion for when cosmological interpretations are warranted.