My guess is that this, in the absence of the acoustic barrier outlined in the first post (i.e., an effect you would use in a car) , is simply using a time delayed and inverted version of the opposite speaker to cancel crosstalk from said speaker at the "near" listener's ear, relying on left/right bias between the ears to separate the cancellation wave from the desireable (left speaker at the left ear, right speaker at the right ear) wave...In any case, is there a mechanism for dealing with comb filtering/secondary crosstalk created by the action of the plugin itself? It seems to me that the fundamental difficulty with stereo is still present...i.e., any out-of-phase/time delayed information you create to cancel crosstalk from the opposite speaker, in itself creates anomalies (comb filtering, etc...) at the same time that it works on the crosstalk. It is beyond my ken how to make this fully effective using processing alone, sans physical barrier separating the speakers, without serious compromises to whatever tonal balance you had before you began, and to anomalous effect overall. I guess it would come down to whether such effect was pleasing overall, but you end up, no matter how much you tweak it, painting with a very broad brush using this type of processing in a car stereo environment.
It seems that no matter where you turn with stereo in a car, you have to deal with such broad "throw it at the wall and see if it sticks" techniques to recover depth information and room acoustics (this stuff, time delayed and band-limited rear fill, etc.)