I am open to the idea that "boring duplicated regions" performance some vague function through spacing. Some folks have proposed doing experiments where the spacers are removed, or replaced with other sequences, but they are extremely hard experiments to properly do (in a way that convinces the field).

We already know that enhancers "work at a distance" and it's not clear what "distance" exactly means, and it gets into complicated 3D structure of the genome inside a cell; see https://en.wikipedia.org/wiki/Enhancer_(genetics)

Personally I think that the best way to think about the genome is to unlearn most of the preconceptions you learned in genetics and instead think about it in terms of biophysics and development and machine learning: you'll never realyl be able to understand the true function of every little bit, but you cvan probably create an approximate model that explains the vast majority of biology with relatively few variables, and some deep models that contain all the necessary statistics to model these systems accurately.

It sounds like because there is a very complex 3D structure that the 'spacing' function could actually be extremely important. Far more so than zeros in machine code.