Removing one HARPS night changes the HD 20794 contrast variance ratio from 0.009 to 5.00

We test time ordering within nights in public velocities of HD 20794. We compare a last-half minus first-half contrast with an even-position minus odd-position contrast. These have equal expected variance for independent zero-mean errors, including unequal error variances. In 361 selected HARPS nights, one night supplies 99.9957 percent of alternating-contrast power. Removing it changes the ratio of mean squared contrasts from 0.00924 to 5.001, with a 95 percent night bootstrap interval of 2.264 to 11.453. A cut on within-night formal-error variation gives a ratio of 4.985. Two extreme velocities have large reported errors. In 57 much shorter ESPRESSO sequences, two reductions give ratios of 1.463 and 1.524, with intervals that include one. Shared spectra prevent independent confirmation between reductions. HARPS uses separate observations but a different time window. A circular-orbit slope bound excludes the three adopted smooth planet signals as the sole source of the ESPRESSO contrasts. No new planet, stellar process, or instrument fault is established. The bundle supplies frozen exposures, balanced-contrast code, sampling responses, permutations, influence checks, source hashes, figures, and offline steps.

astronomy · HD 20794 · HARPS · ESPRESSO · radial velocity · time ordering · data quality · correlated noise · null planet result