Complete orthogonal predictive coordinates alone do not bind a latent direction to a named intervention. We introduce a mathematically grounded synthetic audit inspired by vessel‑device suit‑cordance. Capacity‑matched least‑squares predictors are exactly equivalent under a full fixed‑output transform, while an anchor‑only observer can recover interpretations only within the span of known perturbation signatures. Six three‑dimensional configurations across 64 seeds yielded a maximum paired prediction discrepancy of $6.7\times10^{-15}$ but a median untransported edit error of $1.513$. Transporting the coordinates eliminated that error. Noisy or weak anchors limited calibration stability, and changing the representation basis required recalibration or verified transport. Across 256 additional fits in dimensions 3‑24, prediction equivalence persisted within $4.0\times10^{-15}$. We then evaluated nine deliberate runnable fault classes over 64 seeds each. All 576 faulty executions completed, yet each violated at least one reconstruction, prediction, delivered‑edit, or scope contract; all 320 valid control records passed. Repeating a faulty implementation produced exact self‑agreement despite disagreement with an independently computed simulator expectation. For an omitted‑direction defect, probe coverage followed its analytic law, and rank‑aware abstention protected unsupported interpretations. Scalar‑noise experiments exposed both missed weak faults and excessive rejection under narrow relative tolerances. These controls provide an executable separation of prediction, semantic support, and scientific acceptance. They are synthetic numerical audits, not clinical validation, neural JEPA‑Anything replication, agent learning, or patient treatment‑effect estimation.
Review