I think early scores are just more wildly variable. I think some kids may test low because even with great "g" (whatever that is

), the kid is too jumpy, unfocused, and distractible to give the best performance. But similarly, a young kid who is really great at focus and paying attention may score high since many 4 yo couldn't do that in the test situation and all these scores are comparisons to norms for age. Since most 9 yo can focus for the duration of a test and understand to pay attention and take it seriously, there is less variation from focus and the score better reflects what the test purports to measure. But a brilliant 4 yo who is focused on the next snack and the pretty tree out the window may not show the brilliance at 4, but may show it at 8.
FWIW, I've seen a lot more scores that dropped from preschool age to school age than stable ones or ones that got higher later. I wish there was a good study actually looking at this, but I've never come across one.
Momx2, the stability of your DS scores across time and tests seems to validate both his PG-ness and the tests themselves. With that stability, I could almost believe they are measuring something real!