DXOMark scores are resolution dependent as well as voodoo dependent.
Ah, there's the crux of the argument. DxO uses voodoo. It requires evidence, though. Care to rephrase that?
The scores are higher for full frame,
Yes they are. In most instances, a lot higher.
but in most cases those differences are only noticeable for low-light/high-ISO shooting.
Or maybe in large prints? The ISO setting has nothing to do with the lens performance, but you have a grain of truth here, if you care to be more specific. (And you can find the info in the DxO reviews.

)
Did I not already mention (twice) that full-frame lenses are usually pretty good with APS-C sensors? But they're even better on full-frame sensors.