Complete short noisy protocol

All 128 software checks pass, including twelve complete native fitting and forecast paths. These cover all three generating models, both processing assumptions and both input-noise formats.

Each two-second example has only six earlier receiver positions and allows eight fitting evaluations. All twelve raw and error-conditioned forecasts meet the physical error screen, but none passes the numerical stopping checks, and none uniquely identifies the later prediction through unrestricted calibration. Those outcomes are preserved. This is an interface qualification, not successful long-duration prediction validation.

The calculation now distinguishes two questions: predicting another receiver reading includes receiver errors; predicting the aircraft's physical position does not. The checks preserve identical training covariance, remove only future receiver contributions for the physical target, and save both forecasts before opening later truth. Withheld positions never choose weights or corrections.

The receipt verifies 146 artifacts, including 137 source snapshots, native compilation, twelve charged starts and all work. The test attempts 11,352 native calls and 108,612 normal variates; it reads no aircraft or protected confirmation measurements.

The shared evaluator and saved-results auditor also pass the same 128-check matrix; this is software verification, not an additional set of physical observations.

Next is a separately frozen, longer noisy fitting study with more training positions and the existing numerical and physical acceptance checks. The original failed twenty-minute noisy study and 53-stretch findings are unchanged.

Frozen short-case plan · Nonlinear validation plan · Next analyses.