Joint fits retain the HD 111232 19.7-day amplitude across tested noise models in 26 ESPRESSO nights

We test how the reported 19.679-day HD 111232 velocity curve depends on the noise model. We use the same 26 public ESPRESSO nights from 2024 and 2025 as the earlier SCI test. We add no new observations. We fix the period, phase, and eccentric shape from the original HARPS search. We fit a signed amplitude scale. Separate quadratic or cubic seasonal curves remove slow velocity changes without fixed giant-planet orbits. We test white, exponential, Matern, and quasi-periodic Gaussian noise. The controls add processing-version offsets, three linear line-shape terms, and a wider rotation range of 10 to 100 days. When signal and noise are jointly fitted, the scale ranges from 0.914 to 1.005 across these models. With cubic curves and white noise, the refitted likelihood statistic is 32.14. It falls to 16.54 with long-lived quasi-periodic noise and processing offsets. In the latter model, a scale profile has a likelihood support set of 0.713 to 1.140. This set permits twice the log-likelihood loss to be at most 3.84. This set has no calibrated confidence coverage. Fixing the covariance from the no-signal fit instead gives a scale of 0.297 with conditional standard error 0.191. Fixing the covariance from the signal fit gives 0.914 with conditional standard error 0.081. Thus, stable joint-fit amplitudes and small conditional errors do not establish a planetary origin. We preserve verified ESO headers, source provenance, code, fit audits, likelihood profiles, and conditional null simulations. The work supports further study of the existing candidate. It does not confirm a new planet.

HD 111232 · ESPRESSO · HARPS-RVBank · radial velocity · planet candidate · stellar activity · Gaussian process · noise model · likelihood profile · reproducibility