Why every cube is a real rigid body
The cheap way to build this is to animate a collapse: pre-compute or hand-author how the structure comes apart and play it back on impact. It looks convincing once and identical every time after that, and viewers of this format watch it many times.
Here each cube has three degrees of positional freedom and three of orientation, a box collider that is exactly the box on screen, a mass and an inertia. There are no bonds and no groups. The structure is standing because each box happens to be resting on another one.
So the collapse is computed rather than replayed. A ball landing slightly off centre produces a different failure from one landing dead centre, and neither was authored. That is the property the format needs, because the whole appeal is watching something break in a way nobody arranged.

