It is a byproduct, basically a net zero force excitation is superimposed over the regular driving of the motor. This excitation is used to measure the inductance and infer rotor position at low speed.
There is a possibility that a poorly calibrated or not very salient motor could cog, luckily only at low speed in the sensored range but still be careful anytime you test a new setup. As is, my bench testing with a few setups looks fairly robust but more testing needed. I expect as time goes on and we get a wizard to properly assign values it should only get more reliable as well.
Ben has actually mentioned to me that as of testing right now two shunts is optimal for HFI (only because its used at very low speed and low modulation). The most important thing is having a uniform measurement so always the same two shunts are used for it at low speed. So two shunt designs will work as well for HFI (at the time of writing) as other designs but the other tradeoffs with regard to the rest of the control still exist.
Yes a big update to bring the Unity back up to date is coming and this is definitely a feature that will be included.