The first integrated candidate does not pass the full gate
18/24 controls passed. All six broad-domain cases stopped at the initial source's unchanged nonlinear-cap check. The study is complete and will not be retried. It charged 60 native evaluations, 348 complete bounds and 30,144 bound steps in 136.30 seconds. Failed attempts and unused allowance remain preserved.
The successful controls retain the reference centers and all 51 columns. This partial result does not qualify the candidate or permit a prediction comparison. No aircraft fitting occurred.
The post-run cap diagnosis finds that the new proof used a symmetric box larger than the original asymmetric calibration domain. Computing the proof on the actual domain fits the original cap in all six failed cases. The cap and physical bounds remain unchanged. A separate revised candidate must still pass the complete native matrix.