> It's much, much more likely that a lot of those gene changes overlap, meaning combined changes won't affect the IQ any more than either/or changes would.

No, it's not. Please see Hsu's extended comments on the topic of interaction and non-additivity at the top of the page, and note the support from twin studies & GCTA. The point of his suggested strategy is that you're targeting the subset of gene changes in total heritability which don't do that.