Despite _not_ passing --freeze-tracking-on-disconnect, I observed several times my tracking froze momentarily, with all of my trackers being referenced to my HMD position. This was a bit weird as suddenly my entire body swung around with my viewpoint and is not desired behavior when my hip momentarily has poor tracking or reception.

According to Kung, during my session, my tracker(s) briefly disconnected, and IK2.0 stopped updating tracker positions to avoid misassigning tracker indexes. That itself is fine, but because the tracker offsets then become referenced to HMD position, if I turn my head even a little, my feet move along a very wide arc.

I'd suggest that, instead, trackers keep their position relative to the world space position of the HMD as a fixed world space offset, but do not track rotations of the HMD. This would minimize this exaggeration effect.

One could argue that this would result in strange twisting behavior when locomoting around. However, you'll get strange behavior anyway because you'll freeze mid-gait, so it's a matter of choosing what kind of weirdness is best there. When lying down, however, a momentary freeze of IK processing is much less impactful than tracking HMD rotation.