The GJ 176 fit gap near 109 days is exceeded in 0 to 70 of 512 draws on one bank under four Gaussian noise models

We compare one common Kepler orbit with separate HARPS and HIRES orbits near three reported signal periods. A gap is the common weighted fit cost minus the two separate costs. We use 548 public instrument nights for GJ 176, HD 14412, and HD 140538 A. All five orbit parameters can differ between instruments. Each test searches all 36,736 nodes of one fixed orbit bank. Linear coefficients remain free at every node. Four fixed Gaussian noise models use extra independent scatter or an exponential time covariance. For each star and model, we generate 512 data sets at its best common bank fit and repeat the full bank search. For GJ 176 near 109 days, the measured bank gaps are 62.471, 20.473, 15.278, and 25.746. The numbers of simulated gaps at least as large are 0, 5, 70, and 2, respectively. For HD 14412 they are 0, 263, 386, and 320. For HD 140538 A all four counts are zero. A zero count has a two-sided 95% binomial upper bound of 0.00718 for that one generator. Continuous searches and a finer GJ 176 bank provide numerical checks. Ten of twelve common continuous fits reach the eccentricity limit of 0.9. Under a circular inner orbit assumption, the GJ 176 common fits have outer pericenter radii 0.535 to 0.622 times the known inner orbit radius. A later fit with eccentricity at most 0.8 keeps radial separation. Its common cost rises by 1.307 and 0.497 under the two time covariance models. This is not a stability proof. The simulated fractions depend on the assumed noise and one fitted common orbit. They are not uniform null probabilities or selection-adjusted detection probabilities. No planet is confirmed or excluded. The package preserves code, data provenance, complete results, figures, numerical checks, and reproduction instructions.

GJ 176 · HD 14412 · HD 140538 A · HARPS · HIRES · radial velocity · orbit fitting · Gaussian simulation · stellar noise · reproducible research