GJ 3222 HARPS night means give similar Gaussian fits at eccentricities 0.80 and 0.94 with fixed published periods and closest approach dates

We tested the extreme orbit shape reported for GJ 3222 b with 61 public HARPS exposures from 30 nights in RVBank v1. We fixed both published periods and closest approach dates, and the eccentricity of the long orbit. At each short orbit eccentricity we fitted both RV amplitudes and angles, two instrument zero points, and Gaussian extra noise for independent night means. At eccentricities 0.80 and 0.94, negative log likelihoods are 78.8698 and 78.6604. Twice their difference is 0.419. The short orbit amplitudes are 5.43 and 10.13 m/s. Full model predictions differ by 0.293 m/s RMS at measured nights, while the short orbit curves differ by up to 9.79 m/s near closest approach. At the fixed published eccentricity of 0.929, no exposure has true anomaly within 90 degrees of closest approach. That passage lasts 2.876 hours; the nearest exposure is 5.451 hours away. We checked a later data release, individual exposure fits, separate noise terms, and omitted nights. We also allowed limited changes to the short period and date. The fit penalty at eccentricity 0.80 remains small in each check. This conditional audit locates missing observing phases. It does not give a full orbit probability, reject the published orbit, or repeat the joint fit to velocities and astrometry. Source target data, provenance hashes, code, numerical checks, tables, and figures are preserved.

GJ 3222 · HD 21175 · HARPS · radial velocity · eccentricity · phase coverage · reproducible research