> - The variation between IQ tests done by the same individual is significantly higher than almost any reasonable cross-group variation (men vs. woman, country A vs. country B, doctors vs. craftsmen, ...)
I don't know about your other points, but this doesn't seem believable.
Can't find the one I actually had in mind right now, but here is one where they run various cognitive tests on high-IQ-value individuals, with performance regularly dipping below population average: 10.1007/s12207-021-09417-x
Well it does sound reasonable but I don't think it is as convincing as OP thinks.
If you take two IQ tests in your life and get 130 and 138, I imagine the variation to be higher than the group vs group comparisons he mentioned (maybe not doctor vs craftsmen? it is after all very correlated to education level) but it doesn't mean the measure is "nonsensical and ever-chaning values such as whether I liked today's dinner.."
The point of such indicators is to make predictions about the real world. "If you're 5'6'', this mattress is a good fit" or "People taller than 6ft will be 2.4 times more likely to develop back problems".
Now imagine you try do anything intelligence-related, like hiring the best person for a job. If the variation between person A taking your intelligence test in week 1 and week 2 is higher than between person A and person B, you cannot use this metric to decide who is more intelligent — unless you're prepared to run the test dozens of times, and are fine with the "intelligence" of the person you're hiring regularly dipping below and above the "intelligence" of the other person you could've also hired. At that point you can very much go by firmness of handshake instead.
I don't know about your other points, but this doesn't seem believable.
Is there a reference for this?