[sokoloff]

There's strong evidence that early childhood nutrition, how many words of adult conversation a child hears in a week, how safe and secure they feel as infants/toddlers, and many other differences in children's early experiences can cast a much longer shadow over their lives than the difference of being put into a class where they learn single variable algebra in 6th vs 9th grade.

What right do we have to impose these impacts on children? Or is the school tracking a form of trolley problem where it feels wrong to some to take an active choice?

Because you asked the direct question, I'll give a direct answer. "How early is earlier?" Based on an N of 5 (myself, my two siblings, and my two kids), I think it can be done productively as early as 3rd grade or so. (For me, it was done earlier than that [age 5], which I'm extremely happy about, but I'm not sure that's scalable or even typically appropriate.)

I do think that a periodic re-assessment is also appropriate. If someone "runs well" in 3rd grade and then reverts to the mean, they should revert to the mean academic leveling as well. If someone develops late and starts to excel in 7th grade, there should be a way for them to pivot towards the advanced classes then as well.