Well, I think there are two separate but related issues here. First, testing before around age 9 tends to yield variable results for at least a significant subset of children. Second, children actually have varying growth trajectories with the early bloomer and the late bloomer as extreme cases in point.
The problem with an early identification system is that you will end up with a subgroup of children who may have trouble keeping up by 4th or 5th grade as well as a group of unidentified children who are actually more capable than the previously identified gifted.
The students in the GT stand-alone classes at our elementary are not truly gifted as this group includes more than 15% of each grade. However, one way our school has gotten around the mis-identification issue is by having an unofficial GT class starting in 1st grade, with the GT class becoming official only in 3rd grade. Furthermore, even some kids who secured a spot in the official 3rd grade GT class are not officially identified as GT. Having had three children go through this systems, I have seen enough movements in the composition of the classes to justify the school's wait until 3rd grade for official identification. I think maybe around 1/4 to 1/5 of the classes get moved in either direction. Having said all that, I do not have direct experience as all three of my children were in stand-alone GT classes from 1st grade.
Anyhow, in answer to your direct question, it actually is not "nonsensical to think that a child is not gifted in second grade but suddenly would benefit from services when older." This is particularly possible when you consider that we are talking about not testing gifted as opposed to not being gifted.
