Mathematician, not physicist, but it seems reasonable to me that you could encode a sufficiently constrained 3D space on a 2D boundary. And if you have such an encoding, it also doesn't seem insane that some things might be more easily modeled on the 2D boundary than the 3D space.

If you can convert back and forth between a 2D representation and a 3D representation, and different phenomena are more easily modeled in each, does it matter which is "real"? Unless of course you can come up with a specific prediction and experiment to test it.