Three GBM records at similar orbital phases show a 3.20 microhertz frequency deficit in 4U 0115+63 in September 2026

The preliminary Fermi/GBM frequency estimate near 2026 September 9 for 4U 0115+63 lies 3.196 microhertz below a line between the August 16 and October 3 estimates. We use the last three detections in a frozen public monitor snapshot. Their gaps are 23.988 and 23.995 days, close to the 24.317-day binary period. The formal standard deviation is 0.513 microhertz under independent supplied errors. Two central orbit alternatives from Li, Wang and Zhao (2012) change the contrast by less than 0.02 microhertz in an instantaneous approximation. Unknown weights within centered three-day windows give broader approximate ranges, from -5.639 to -0.752 microhertz for the fixed-angle alternative. Four of 31 earlier overlapping triples have absolute contrasts at least as large. Chosen Brownian frequency noise of 0.3 microhertz per square root of day reduces the contrast-to-error ratio from 6.23 to 2.76; this noise is not fitted. The analysis is exploratory. Weak detections, unknown timing noise and the lack of a photon analysis prevent a claim of unusual physical torque. Swift/BAT has no daily rows during the first 16 GBM detections in 2026, so no torque-versus-rate model is fitted. The package preserves source data, hashes, all comparison triples, code, figures, numerical checks and offline reproduction instructions.

4U 0115+63 · Fermi GBM · X-ray pulsars · binary orbit · pulse frequency · timing noise · Swift BAT · reproducible analysis