Four full-simulator timing checks pass

All four original-start fits pass the complete numerical and raw forecast screens. Both matched starting-point comparisons agree, and all four original input, training and error-operator bindings replay byte-for-byte.

The run finishes in 10.7 minutes after 189 source-bound software checks. It uses 238,386 actual native calls and 126 bounded linear operations, within its prepaid allowance. No aircraft observations or protected groups are used.

The earlier midpoint-anchor trial remains failed. This separate version checks both sides using cell midpoints inside the existing ±10-microsecond window. Every nuisance column and acceptance threshold remains unchanged. The two originally failed calibration fits are not relabeled: these are newly frozen regressions from their original starting points and unchanged inputs.

This qualifies four exposed simulator cases, not the whole model/input matrix or aircraft uncertainty. The next gate is a separately frozen 48-case matrix covering both input formats, all candidates and both processing settings. Calibration remains closed. The original 53-stretch findings are unchanged.

All four results · Starting-point comparisons · Frozen plan · Next analyses.