addforceatposition starts off with a wonky change of rotation

I'm trying to do a simple bit of physics here where I'm using dummy objects attached to thrusters on a ship. I'm using AddForceAtPosition-- it seems to work, except, the first frame or two after setting the simulation in motion causes the rigidbody to rotate around the Y axis to a specific but unexpected angle.

This code is being executed on the fixed update:

 MyGimbal.AddForceAtPosition( LThrust.forward * MyProfile.impulseFor, LThrust.position );
 MyGimbal.AddForceAtPosition( RThrust.forward * MyProfile.impulseFor, RThrust.position );

LThrust and RThrust are transforms that are attached to the main object. They are essentially flanking the center of gravity, swept back slightly, with the forward axis parallel to the rigidbody's transform's forward axis.

What's odd, is it works brilliantly except for the part where it immediately flops around the y-axis to 243.3798; oddly specific and consistent.

All I can think of is that bit in the manual where it warns "Note that when position is far away from the center of the rigidbody the applied torque will be unrealistically large."

I like how it doesn't give any qualification of "far" Ultimately, this can't be the problem because it doesn't matter if I set the position to the actual position of the thrusters, or if I scale the offset down.

Changing drag doesn't effect it either.

I also thought maybe the first time the force is applied by one of the thrusters, that it's pivoting the rigidbody around by the 243 degrees or whatever, but if I swap the order, it still happens in the same direction.

What is going on here??