Some comments:
OP mentions IQ, but while IQ is obviously strongly correlated to math ability, they are not the same, so it's better to use math ability tests than IQ tests for math acceleration decisions.
That said, I think similarly to the OP, that a kid (say 8yo) who is at least 2SD above average in math (2.5%=100,000 out of 4,000,000 8yos in the US) should be able to handle a 2 year acceleration in math, and a kid (say 8yo) who is at least 3SD above average in math (0.1%=4,000/4,000,000) should be able to handle a 5 year acceleration in math.
(These numbers seem instinctively right to me, but of course it would be better to have it properly studied.) In any case, large numbers (though small percentages) of students should probably be accelerated a lot, and probably a lot more than is happening now.
What I do know is that students just 1SD above average in math, can achieve 2 or more years of progress more than average students over their K-12 years, simply by going about 20% faster, in an ability grouping setting (without grade skipping) which was the situation when I was at school. The top group, which was about the top one sixth, or 84th percentile and above (standard score 115 and above) reached calculus by about 9th or 10th grade.
(I've got more to say, but I'll submit this so far.)
