Rotation/gain structure helps modestly
All 48 saved cases and 288 independent support checks passed. Complete horizontal ranges narrowed by about 1–3% relative to the vector-norm comparison. Height ranges were unchanged. Three of six model/processing groups still miss the unchanged twofold target on every axis.
The calculation finished in 67.1 seconds with no native integration or bounds, aircraft fits or original-file reads. It preserves the original physical limits, all 51 coordinates and the original sectors. Known truth remains feasible and contained in every applicable sector; predictions precede the truth checks.
More constraints on the same matrix should not be assumed to solve the remaining bottleneck. The next saved-only diagnostic will compare certified upper bounds with exactly verified feasible witnesses in the existing relaxed region. This distinguishes certificate looseness from width permitted by the construction. A feasible relaxed point need not represent a physical calibration.
The aircraft findings remain unchanged. Useful recording-length ranges, whole-program numerical enclosure and observational error coverage remain open.