Also not sure how valid this one is as data packs are still experimental. Found while testing MC-215417.

In the dimension added by the attached data pack, the buildable range is locked between y=-2032 and y=-2016. Despite this, clouds still appear at about y=192, despite this not being visible in the slightest from the actual ground level.

This could be fixed by making clouds depend on the actual y limits themselves, or even better, their height made customizable independently of them in the dimension definition itself.