Mapman, what you say is correct. In many respects, a movement away from statistics and toward first principle understanding and modeling provided a better level of certainty with respect to outcome. However, in early development proceeding with interim manufacture with minimal understanding is quite common. In industries where the attrition rate of developmental products is quite high, it does not make sense to spend that much money until it is clear a product is going to go to market. In my industry, a lack of understanding at a mechanistic level does not represent a potential risk to the customer.
Not only are the stakes lower in audio, but the ability and resources of the average Audiogon poster, be they skeptic or innocent gullible fool, does not permit that level of analysis. I've never seen the sort of thing Sabai is complaining about addressed via a first principles argument. It usually runs more along the lines of you're a stupid idiot, na na na na na na.
I know I won't spend my time pursuing that level of understanding in audio.
Even if such an argument were presented, it does not address the far more unpredictable human response. Assuming the tweak produces even a marginal difference in sound, that difference could be objectively either "better" or "worse," which may or may not correlate to a listener's subjective response to that difference.
I think Jedinite's comment is a good one. Some of these products could have some scientific basis, but if they are marketed behind metaphysical mumbo jumbo, I'm probably not going to bite.