Moisture rate was first-minus-last over 6 h, so a probe reinstall (air 1 % -> soil 68 %) read as +66 %/h ‘wetting’ and forced dryback to 0.00 % for six hours (‘

Operator screenshot 2026-09-14: moisture 67.9 %, DRYBACK 0.00 %, ‘Rate 66.37’, ‘dry-back -0 % since the last wet point’. The rate took the first and last rows in a 6 h window; with the pre-install air reading still inside the window the slope was a single step, and _dryback_pct treats any rate above +0.5 %/h as a watering in progress and returns 0. Fixed in PR #259: Theil-Sen median of pairwise slopes over one sample per minute in the last hour (a single step cannot move the median), tests in tests/test_sensor_trust_history.py. Verify: Rate settles to a small negative number within an hour of the reinstall and DRYBACK starts counting.