Servicing the async commutator without breaking three-point suspension

The async commutator transmits backpressure through a spherical pilot channel. A 60-cycle event source energizes the channel. The field then induces a voltage in the worker tasks. Continuations flow through the commutator to the main dispatch shaft. This is documented. The grounding procedure hangs above every desk.

The precession check is not on the wall.

Service a commutator without the precession check, and it passes every load test. It then transmits backpressure in the wrong direction at the worst possible hour. The tasks do not fail. They resolve toward the dispatch shaft instead of away from it. A task that moves toward the shaft looks like a task that is nearly done. Then it stops looking like that.

$ dispatch-cli precession --channel pilot
channel pilot: 0.4°/hr precession (nominal < 0.1°/hr)
secondary stator tasks: 12 of 14 phase-locked
recommendation: re-torque stator before next deploy

The re-torque takes two minutes. The incident it prevents takes four hours. You spend those hours staring at a trace. Every span is green. The p99 is eleven seconds. One task walks the wrong way around the commutator. It started walking at the last service.

Three rules, in order of how long they take to learn. Ground the channel. Check the precession before the re-torque, never after. A precession reading taken after a re-torque measures the torque, not the channel. The rig sits in emergency stop already? Do not service the commutator. Service the suspension first. A serviced commutator on an ungrounded rig is the technical equivalent of a shrug.