You and the other commenter bring up good points. Developmental neurotoxicity (with lesser or no effects in older children and young adults) is, I speculate, probably due to differential gene expression during early development versus later when genes related to development are suppressed and genes related to maintenance are more abundantly expressed. The developmental neurotoxicity probably works through different mechanisms than what is termed "postoperative cognitive dysfunction" in the elderly after general anesthesia dysfunction [1][2], which, all I know is that it is a thing. If I were to speculate it would be that in the elderly there are fewer redundant cognitive resources, and so detrimental effects to cognition are magnified. I know that it used to be thought that post-operative dysfunction is temporary, but it seems likely to me (again speculation) that there is both recovery and permanent dysfunction, but the dysfunction becomes a little more difficult to detect. Going back to my paper, where we used a method to disentangle two types of memory processes i.e. recollection (explicit recollection of experiential details) and familiarity (a general feeling of familiarity with things you've seen previously) which contribute to memory performance but tend to be differentially affected by neurodegeneration (recollection is more affected, and generally more hippocampal), so that sometimes, when not accounting for these processes, a memory test will fail to find differences because patients rely on familiarity to answer memory questions.
[1] https://scholar.google.com/scholar?as_ylo=2020&q=postoperati... [2] https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&as_ylo...