The issue isn't some innate flaw with intelligence tests. The problem is testing kids too young, with any kind of test.
The real problem:
In applying the science to the reality, the problem doesn't seem to lie with the age of initial screening. Even in kindergarten, a few children are clearly and indisputably advanced. Instead, what stands out as problems are: the districts who don't give late-blooming children additional chances to test in, and the lack of objective retesting to ensure the kids who got in young really belong there.
What everyone should know:
[I]t needs to be recognized that no current test or teacher ratings systems, whether used alone or in combination on such young kids, meets a reasonable standard of confidence to justify a long-term decision. Huge numbers of great kids simply can't be "discovered" so young.
How the tests really fail our kids:
Real intellectual development doesn't fit into nicely rounded bell curves. It's filled with sharp spikes in growth and rough setbacks that have to be overcome.
We need to question why this idea of picking the smart children early even appeals to us. We set this system up to make sure natural talent is discovered and nurtured. Instead, the system is failing a majority of the kids, and a lot of natural talent is being screened out.
From Chapter Five: The Search for Intelligent Life in Kindergarten in NurtureShock: New Thinking About Children by Po Bronson and Ashley Merryman