Does removing anti-skating really improve sound?

Mijo, No.
I was going to leave it at that, but let me take your statements in reverse order:

"Tracking angle error has very little to do with it as long as the arm is set up correctly. It is all about friction and offset angle and offset angle is fixed."  It's actually about the friction force, we both agree, and about the degree to which the cantilever is not tangent to the groove.  A major reason for lack of tangency, for all tonearms that are mounted such that the stylus overhangs the spindle, is headshell offset angle.  Again, we agree.  But in addition to the headshell offset angle, there is a variable lack of tangency to the groove, created by the tonearm geometry.  At the two null points of any typical alignment algorithm, where the cantilever is tangent to the groove (tracking angle error = 0), there is still a skating force due to headshell offset angle.  But at all other points on the LP surface, headshell offset angle is not the only cause of lack of tangency.  That additional ever changing angle is one reason why the skating force is changing all the time.

"The more tortuous the higher the velocity. The force required to move the stylus back and forth is seen by the system as friction. So, the higher the groove velocity the higher the friction."  That's a very appealing concept, and I was sort of thinking that way too, but "velocity" does not appear in the equation for friction.  So you and I cannot have it that way. The way I see it now, the tortuosity of the groove results in those mini-acceleration events; the stylus MUST obey the speed of the turntable.  No matter how many twists and turns in the groove.  Thus, the stylus is experiencing mini-accelerations (acceleration = change in velocity) as it is dragged around the LP at a constant net speed through those zigs and zags of the groove.  Acceleration does generate a Force.  (Remember, F = ma.)  Those small forces being generated, acting through lack of tangency to the groove wall, must be resisted by the cantilever, the cartridge body, and ultimately the stiff tonearm to which it's attached.  That contributes to the skating force as well.

Lewn Theoretically you are correct in that varying tangency will change the force vector but in real life the error is so small it does not account for much. 
You are mistaking velocity with record speed. As the modulation increases so does velocity. In a heavily modulated groove the record has to exert more force on the stylus to move it and the stylus is moving faster. In a straight groove the record is applying only straight ahead friction and the stylus is not moving (through space) at all. The energy (work) required to move the stylus is seen as friction and a little heat. You could make the argument that anti skate does not decrease towards the center of the record because actual velocity does not change. There are just as many squiggles they are just closer together. And indeed it does not change as much as you would expect given the vast difference in groove speed between the outside and inside of the record. I did make a miss statement earlier when I said velocity decreased towards the inside of the record. It does not. Groove speed decreases towards the inside of the record. What would be nice is a test record with a bias band an entire side. That would be fun!

I haven’t read through this thread but here’s my take...

On my VPI, (we know Harry doesn’t advocate anti-skate) I haven’t noticed any difference in sound with or without. It just makes perfect sense to use it for the health of my records. As long as your azimuth is good, that’s much more important for SQ.

On my VPI, (we know Harry doesn’t advocate anti-skate)

From the VPI owners manual

As mentioned earlier, the arm wire applies the anti-skating force. The degree of force applied can be adjusted, to increase anti-skating force give the connector a counterclockwise twist, unwinding the wires natural twist. Likewise, to decrease the force, give the connector a clockwise twist, winding the wires natural twist. Remember, the Lemo connector can only be "adjusted" in increments of whole turns. If it is not, its key will not line up with the groove in the receptacle