If it's such an objectively verifiable problem, why do some (and that's Harry Weisfeld included) don't hear any improvement with anti-skate applied?
It's not that we don't hear any improvement. I do. It's just that we hear many more detriments that swamp the improvement.
***
The problem, as I've posted several times over the years, is that real-world A/S mechanisms apply lateral bias to the TONEARM, yet the skating force they're trying to counteract is generated at the STYLUS.
Imagine, if you like, grabbing the STYLUS with your left hand and pulling it inward (skating) whilst at the same time grabbing the TONEARM with your right hand and pulling it outward (anti-skating). This is what's actually happening with skating forces vs. anti-skating mechanisms.
It's easy to visualize that these unequally applied biases necessarily pressure the cantilever against the suspension. This pre-dampens its freedom to make excursions based on groove modulations. Result: softened micro-dynamics, slowed transients, dampening of the finest, lowest-level sounds in the groove. Sound familiar?
This is why excessive (any) A/S sounds almost exactly like excessive VTF. Both pre-dampen the cantilever against the elastic suspension, reducing its freedom.
The ideal A/S mechanism would operate like this: your left hand pulling the STYLUS inward (skating) whilst your right hand pulls the STYLUS outward with exactly the same (ever-changing) force, with zero lag time of course (anti-skating). This would avoid pre-dampening the cantilever and, if perfectly implemented, would carry no sonic penalty.
Of course no one has or ever will build an A/S mechanism based on a perfectly reactive string tied to the stylus and pulling outward. ;-) The mechanisms it's actually possible to build are necessarily imperfect, as described above, and will always carry the associated sonic penalties as well as benefits.