A phase scan over five widths gives p = 0.036 for 94 Proxima observing nights under random event labels

We tested the phase cluster of published Proxima d emission-event flags with the observing times held fixed. We matched 28 event marks from a public figure to a 114-row public timing table. We removed the first three dates and grouped spectra by night. This gave 94 nights and 25 flagged nights. At a fixed period of 5.12338 d, the published 0.90 d window contains 11 flags among 18 nights. Its exact one-sided conditional probability is 0.000610 under random event labels. A scan over all phase centers gives p = 0.0162. A scan over centers and five full widths from 0.45 to 1.35 d gives p = 0.0360, with a 95% simulation interval of 0.0352–0.0369 from 200,000 draws. The latter probability is 0.0367 when each season keeps its own event count. A complete test of 2160 cyclic shifts within seasons gives 0.0255. An exploratory shuffle within season and spectra per night gives 0.0323. Night-time and sample choices give further sensitivity. The finite phase search weakens the occurrence evidence in this public subset. The cluster remains suggestive under the tested null models. The study does not recover the complete published sample, redetect events, or test all reported evidence for magnetic interaction. We provide event matches, source provenance, code, validation checks, results, and figures.

Proxima Centauri · Proxima d · star planet interaction · stellar flares · ESPRESSO · phase scan · randomization test · public data · reproducibility