Evidence for the 15.196-day signal in 2025–2026 BAT data from 4U 1909+07 depends on the noise model

We test whether the known 15.196-day X-ray signal in 4U 1909+07 has returned after the last date used in a 2025 study. We use a saved public Swift/BAT snapshot with 359 accepted daily measurements from 2025 January 7 through 2026 October 20. The selected calendar range is MJD 60671 to 61333. A weighted fit includes a trend, two binary orbital harmonics, and two harmonics at the fixed long period. The fitted long-period root mean square (RMS) amplitude is 0.000605 count cm^-2 s^-1. A block wild bootstrap gives null probabilities of 0.0108, 0.0371, and 0.1054 for 15-, 30-, and 60-day noise blocks. Equal weights give 0.3322 with 30-day blocks. A nearby-period search over 14.5–16 days gives a band probability of 0.0613. The nominal 95% RMS upper bound is 0.001109 count cm^-2 s^-1 with 30-day blocks. This bound depends on the residual-noise assumptions. In 2,400 synthetic trials at the later times, a test that recalculates its null threshold detects the earlier strong-state waveform in 24.2% of trials. This fraction is conditional on the saved 30-day residual model and a fixed earlier waveform. The evidence does not establish a return or loss of the physical cycle. The contribution gives a later-data test, a conditional amplitude bound, and a measured sensitivity limit. The package preserves source bytes, provenance, code, selection records, bootstrap draws and simulation trials, tables, figures, and offline reproduction steps.

4U 1909+07 · 4U 1908+075 · Swift BAT · X-ray binaries · superorbital modulation · correlated noise · block wild bootstrap · injection recovery · public data