BaCaRoZzo

Ok, I've solved my doubts and problems about the inertia tensor and, most importantly, correctly calculated it.
Basically, V-Clip returns the "second moment of area" which is also called inertia tensor. Such matrix can be converted to the (mass) inertia tensor simply by multiplying it for the factor "m / V" with "m" the mass and "V" the volume of the shape.

Thanks to this correction the velocities updated by the response are at least not over the overall initial momentum but the results still seem incorrect.

Before plaguing you with a long post (which would end unanswered like the previous one) I'll repost the same - stupid - question: since at the denominator of the impulse magnitude appears the offset of the shared contact point from the centers of mass, choosing the wrong contact point would end in a wrong response? I'm still thinking to limit cases such as parallel faces in which any point is exactly at the same (minimal) distance.

Any help is really appreciated. Thanks in advance.