Full-calibration controls for horizontal-only refitting

Scope and evidence distinction

The complete six-parameter local-probe study passed all twelve horizontal fits and all six horizontal paired-start comparisons. Its height-only mode remains unqualified: one objective-repeatability comparison still fails after a separate larger-budget continuation. Preserve every failed global/height gate.

This new software capability is explicitly horizontal-only. A saved-only versioned review must verify every predeclared horizontal case, without choosing lower-cost historical fits, and must label height-only capability unqualified. No threshold is changed. This is not a new scientific eligibility policy or a claim that an unqualified coordinate mode has passed.

Fixed new matrix

Before any observed refit, test all 26/27 calibration columns on independent generated motion: three physical models × fixed-zero/profiled upstream removal × zero/opposed starts = twelve generated fits and six paired comparisons. Use nominal increments, the original instrument_limits(1.) bounds, and the exact truth/start definitions in the earlier full-calibration preparation. Score only horizontal receiver coordinates. Heights still enter initialization, covariance and propagation; this is not altitude or datum independence.

Use the separately qualified local-probe/curvature-corrected candidate with unchanged numerical tolerances. Require every fit and paired-start comparison to pass under the preceding objective/error/repeatability limits. Retain all columns, report lost rank and unscored errors, and make no unique-calibration claim.

Finite allowance and continuation

• Twelve starts, each at most 160 tangent calls including fresh checks, plus two diagnostic predictions: 1,944 generated native calls maximum.
• Four total numerical threads, thirty minutes, no retries or expansion.
• Zero original-recording reads and zero observed-data starts.
• Freeze sources, environment, horizontal capability review, exact identities, software checks and allowance before launch. Use locks, pre-call journals, atomic outputs and hashes. Never restart charged or completed work.

The broader 24-case full-calibration preparation remains unstarted; none of its unused allowance is borrowed. A passing horizontal-only stage permits preparation of a small matched horizontal-objective observed study with its own bounds, starts and finite allowance. It does not qualify height-only refitting, calibrate significance or alter the completed 53/0/0 archival comparison.