Dump: CD-08 · Cell R2 (anonymized)
Claim on radio: “Just a bump — jog clear.”
Format: Annotated collision dump. Not a safety slogan.
Dump header
CD-08
Mode: AUTO
Program: PICK_MIX_07
Axis at stop: J3 / J5 peak
TCP speed at detect: 0.42 m/s
Collision sensitivity: level 3 (unchanged 14 months)
Contact estimate: high (tool side)
Stop category: protective stop

If sensitivity is archaeology, every dump is a surprise.
Annotation (what the dump actually said)
| Signal | Reading | Interpretation | | --- | --- | --- | | Path ID | node 18→19 | Same node as last near-miss sticky | | Offset active | +4 mm Z (manual) | World model drifted; not in change log | | EOAT ID | GR-B | Finger chip known from photo log — ignored | | Fence | closed | Not a human intrusion story |
So: geometry + offset + worn EOAT. Not “operator error” as first line.
Forbidden responses (written on the dump print)
- Raise collision threshold to “get through the shift.”
- Delete dump to “clean the log.”
- Resume AUTO without fixture / offset / EOAT check.
- Blame “random disturbance” without node ID.
Clearance gate after CD-08
- Archive dump + video if available.
- Measure fixture vs taught points at node 18–19.
- Clear or MoC the Z offset.
- EOAT inspect (photo log rules).
- Dry run at reduced speed; then restore AUTO.

Jogging clear is not root cause. It is scene contamination.
Adjacent fences
Cycle waterfalls own lost seconds. Gripper photo logs own wear. Payload logs own mass lies. Muting protocols own scanner bypass culture. CD-08 owns collision as a forensic file. Do not tune sensitivity to hide a bad offset.
Close
A bump is a story. A dump is evidence. CD-08 was evidence—until someone almost filed it under folklore.
