The median Cen X-3 reversal duration difference changes sign between 60-day and 100-day isolation cuts in ramp fits to the 2008–2026 GBM history

We test a stated measure of spin-reversal duration in Cen X-3. We use 2,506 public Fermi Gamma-ray Burst Monitor frequency measurements from MJD 54690.4689 to 61307.9979. We find extrema with local quadratic fits. We fit a frequency curve whose slope changes linearly over a bounded duration. A 100-day event isolation cut selects one transition into spin-down and four into spin-up. The fitted median duration difference is +27.4 days. An 80-day cut selects six events in each direction and gives −2.0 days. A 60-day cut selects eight and 12 events and gives −27.6 days. The 60-day sample was added after the small strict sample was known, so its analysis is exploratory. Its 95 percent event bootstrap interval is −59.2 to +32.0 days. Formal-error weights change this sample's contrast to +11.0 days. Six fitted widths are below the local median cadence, and four reach a width bound. These estimates depend on the event and fit rules. They do not establish a population duration difference or a physical disk mechanism. The package includes the exact source snapshot, all candidates and rejection reasons, fitted events, and control results. It also includes noise simulations, figures, code, and reproduction instructions. The earlier visual event selection was not matched. The event rule and duration measure in this work differ from that comparison.

Cen X-3 · accreting pulsars · torque reversals · Fermi GBM · event selection · duration estimation · reproducible astronomy