Prediction calculation: improvement, but not qualification
The study finished in 81.3 minutes. 32 of 48 calibration-region calculations passed; 16 reached the existing per-bound time limit. All failures, charges and partial artifacts are preserved. The completed study will not be retried.
Three of the six model/processing groups have every region available. For the two specified-field globe groups, combining the new and earlier valid ranges reduces height widths by 25–28%. Horizontal widths change by only 0–2.5% and still miss the original target. The complete zero-field/zero-removal globe group passes that target. The other three groups remain incomplete.
The earlier qualified comparison still passes four of six groups. This study does not replace missing regions with successful ones or turn partial ranges into complete results. It adds useful height bounds, but does not qualify the full method for withheld aircraft predictions.
These checks use six generated 7.2-second histories. They add no aircraft fits and leave the 53-stretch findings unchanged. Numerical error bounds, full-recording performance and calibrated observational uncertainty remain separate requirements.
Next: identify the costly propagation operations and examine the remaining horizontal error terms from saved artifacts before changing the method. Keep the physical bounds, failed results and prediction target unchanged.