Well, there may be ways to hack parts of it in, but it's not elegant. Distance-based collisions simply treat the object like a circle, regardless of its actual shape. And what about per-pixel collisions? Supporting that would be extremely difficult and probably prohibitively slow with 3D meshes. So again even a seemingly simple solution is riddled with difficulties and incomplete support. Basically, without looking at the codebase, you can't imagine the immense complexity that goes in to something as simple as checking if two objects overlap with per-pixel accuracy. Then there's the rest of the engine, other areas which are even more complex and present even greater conundrums.
We're not trying to be stubborn here. People are throwing out what they think are simple ideas, but are actual massive technical challenges because they don't know the codebase. The fact remains that even seeminly simple 3D support is extremely time consuming and complicated to develop. Just look at the number of bugs and problems associated with the 3D box plugin, and I can tell you, that alone has taken many, many hours of development time too. It's not just a 3D box. It's a 3D participant in the Construct engine, which is much, much more complicated than simply just being a 3D box.