I think you should in theory be able to match two "shifting" lenses by applying the same degree of shift.

If for tilting you mean exploiting the Scheimpflug law, then I don't see it very feasible as the two lenses would have to be tilted by different angles in order to cover, more or less, the same field of view, that I suppose would mean that the brain would not be able to "reconstruct" the image without some headache as the two images would be different.

Besides, I suppose a 3d camera would be optimized for a certain focus distance (let's say: 20 metres, or infinite). If you try to obtain a 3d image of a near object, the two lenses should be mounted so that they converge (just like your eyes do when you observe a pencil that you move toward your nose). Tridimensionality of non-far objects becomes, thus, quite a complicated matter, and one might expect that you want to add tilt to your images because you intend to include some near objects in the frame.