Found 2026-09-14 from a dev-bed LLM loop round: pot2 moisture railed to 0 % for 270 virtual min under the seeded plant “NL clone A”. probe2_reading_dark came on, but probe2_not_reporting never did. Cause: soil_tests.init_probe_station_defaults marks pot2 (2×4) and pot4 (4×8) as probe stations (reading_mode idle) on first startup of every install. Assigning a plant to that probe (plant_probe.assign_plant_to_probe) sets assigned_plant_id but leaves the station role in place. sensor_trust then treats the probe as a parked station: the flatline dark path never alerts, and probe{n}_sensor_stuck / untrusted skip the flatline check entirely. An operator who plants on probe 2 or 4 silently loses dead-probe and stuck detection for that plant. Peer MAD already treats ‘station OR no plant’ as excluded, so a station is really parked only while no plant is assigned. Expected: exempt only a station with no assigned plant.
Journey event · 14 Sep 2026
A probe station with a plant on it is treated as parked: a dead or railed probe under a live plant raises no not-reporting alert and skips stuck detection (pot2
Found 2026-09-14 from a dev-bed LLM loop round: pot2 moisture railed to 0 % for 270 virtual min under the seeded plant "NL clone A". probe2_reading_dark came on, but probe2_not_rep
● state copied from the tracker, never improved
● event date 2026-09-14
| Event date | 2026-09-14 |
| Version found | |
| Version fixed | |
| Expected kit firmware | |
| Release file | |
| SHA-256 | |
| Served count | |
| Source | https://app.notion.com/3da2b4cda370815ca011e209002a537b |
Empty rows are honest
A blank cell means the tracker has no value for it. Nothing on this page is filled in to look complete.
