All six native regressions pass

All six fits, three starting-point comparisons and six original-input replays pass the independent audit. The run finishes in 10.6 minutes using four workers with one numerical thread each. Every fit passes all nine timing-cell checks under the original numerical and position-error limits.

The cases include an earlier successful control, an earlier numerical failure and the latest calibration failure, each with its two prescribed starts. The first two retain all 27 nuisance parameters; the last retains all 26. Every input, training observation and error-operator binding matches its original. No earlier result is relabeled.

For the latest formerly failing case, the corrected verification accepts the fit at the same fitted cost and prediction errors: about 0.70 m horizontally and 1.60 m vertically. This checks the numerical correction without changing the simulated measurements or scientific acceptance thresholds.

The six fits use 220 evaluations and 154 linear operations. Actual native work is 357,544 calls and 435,412,458 packet visits; the larger prepayments remain charged. There are no retries. The execution passed 338 source-bound software checks before freezing. A read-only auditor metadata repair subsequently passes 339 checks. Its source, regression test and binding record are included in the publication receipt; no fit was rerun for that repair.

These are exposed generated-data regressions, not new aircraft observations or independent statistical validation. The next gate is a fresh 48-case native matrix covering all three generating candidates, both input formats and both processing settings. Calibration remains closed without a usable threshold, and the 53-stretch aircraft findings are unchanged.

Verified report · All fits · Starting-point comparisons · Frozen plan · Next analyses.