Looks like this might be limited to spindle speed and direction pins not consistently reacting to a program stop as 'spindle.0.on' seems to always clear.
Good catch. I will look at these pins in case it happens again. I had it only a few times when milling PCBs. Other than that it never occurred...