thanks for the update!
so does the aliasing happen at an intermediate stage where a sample rate conversion is done to save computation cycles in subsequent code? if this is the case, wouldn't it be cheaper to use a more aggressive anti-aliasing filter before down-sampling than to stay at the higher rate for functions that don't require it?
if not, i like the idea of making this optional. there are too many who have told us we aren't hearing it, or it's our poor technique, or it doesn't matter, who will likely explode if they don't get their prized new feature because of this. otoh, this will take additional coding effort as well.
i would rather see a fix for this than an attempt to get every feature under the sun in one box. if you need one box to rule them all and in the darkness bind them, there is already an option...