The ExoPdot v1.0.0 rule gives half the Gaussian covariance for linear and quadratic transit clocks at convergence with fixed errors and inactive bounds

This study checks the uncertainty calculation in the fixed ExoPdot v1.0.0 archive. The linear and quadratic transit models use exp(chi2_old - chi2_new) as a Metropolis acceptance rule. The Gaussian rule needs a factor of one half in this exponent. With fixed independent Gaussian errors, flat priors, symmetric proposals, and inactive bounds, the source rule gives half the correct parameter covariance at convergence. An exact least-squares audit covers all ten archived cleaned-target results, with 20 fits and 50 parameter intervals. The median saved 16-84 percent halfwidth is 0.707213 times the correct Gaussian halfwidth. The predicted ratio is 0.707107; the observed range is 0.681802-0.711859. The package includes a two-line correction and checks of the original source functions. A second application measures ten HAT-P-19 b transits from public TESS flux, including four in October 2024. The later sector mean exceeds the early quadratic forecast by 138.15 seconds. Its conditional score falls from 3.21 to 2.05 when the early covariance includes excess scatter. Other timing-error choices weaken this test further. These data do not give a decisive result for physical orbital decay. The source audit does not test the nonlinear precession model or repeat the original clipping. It does not imply that all fitted period trends are false. The publication preserves input bytes, hashes, source code, full covariance results, figures, and offline instructions. A fresh-directory run reproduces all 43 result files.

transit timing · ExoPdot · Metropolis sampling · Gaussian likelihood · covariance · orbital period · HAT-P-19 b · TESS · source code audit · reproducible research