This may be the same bug as #2880 but it might not be.

A user in the KSP forum reported a problem and gave an example save file that showed it.

(Link to thread: https://forum.kerbalspaceprogram.com/index.php?/topic/165628-ksp-1101-and-111-kos-v1310-kos-scriptable-autopilot-system/&do=findComment&comment=3948357)

With this design of vessel, the steeringmanager claimed the torque the vessel is capable of is about 2.5 kN*m when a back of the envelope calculation seems to show it should be more like about 1 kN*m. Being this far off in how much torque it thinks it has causes oversteering past the target direction and wavering back and forth.

I think something may be wrong inside SteeringManger.UpdateTorque() making it give a bogus value. Note: this is a vessel that only had RCS thrusters for torque and no reaction wheels, so that means it is using kOS"s homebrewed ModuleRCS torque calculations, so there may be a flaw inside that logic.