A fixed 6/7 BAT law with frequency random walk noise leaves correlated spin residuals in 2023 data for Swift J0243.6+6124

We tested a relation between hard X-ray rate and spin change in Swift J0243.6+6124. We used frozen public Swift/BAT and Fermi/GBM data from the 2018 decay and the 2023 outburst. The selection retained 42 and 29 frequency changes with interpolated BAT rates between 0.03 and 0.30 count cm^-2 s^-1. The mean used a fixed rate power of 6/7. The covariance included shared frequency errors and a Gaussian frequency random walk. Even with separate mean relations, the 2023 whitened residuals had a lag-one correlation of 0.568. None of 5,000 refitted simulations gave a value this large. The 2018 value was 0.039. The correlation remained positive with free orbital harmonics, BAT error cuts, and alternative pointing averages. The primary slope ratio was 1.070. It changed to 0.874 or 0.590 when only the 2023 torque noise had a fixed correlation time of 3 or 7 days. The primary noise failure prevents a firm physical torque constraint from its nominal simulation range. These exploratory checks identify a failure of the tested model and input treatment. They do not select a physical cause. The package supplies frozen inputs, source hashes, code, simulation tables, figures, and reproduction instructions.

Swift J0243.6+6124 · accreting pulsar · Fermi GBM · Swift BAT · spin frequency · torque noise · model check · correlated residuals · reproducible analysis