Yes, it seems to be working in conjunction with the torque vectoring to keep the car as stable as possible. If all the regen was done on the rear motor it wouldn't be as good in some situations. I did similar testing to Bjorn on a dry airfield and in those conditions it was consistently a 25/75 split. I also tried it on a skid pan but don't have the data to hand for that test. The main thing is that whatever you try and do to the car to unsettle it, 99 times out of 100 it can sort it out for you with a combination of regen and braking.Bjorns video certainly shows it shuffling the regen around to wherever it thinks is best depending on grip. You can see each motors torque with the scanmytesla recording, sometimes its entirely the front motor doing the regen, sometimes the rear, and sometimes a split.
4:53 in particular shows it on snow moving the regen towards the front as there's not enough grip at the rear