> Does anyone defend this as being useful for anything?

Extreme relativistic phenomena like "spooky action at a distance" or the Alcubierre's warp engine folding space may be intuitively easier to grasp if you think of space as a structure made of wobbly interacting elements approaching or receding from one another, rather than extreme deformations of abstract euclidean geometry or probabilistic waveforms.

For example, time ticking at different rates in different places could be explained by a constant rate of causal propagation through regions with a different density of these elements. Or, alternatively, by causation propagating at different rates going slower through elements with very high energy. This can create two competing models, which could be tested experimentally to see which one is closer to reality. (This capability to ultimately being tested empirically is the main thing that differentiates theoretical physics from scholastic speculation; if not for that, both would be rationalistic masturbations to see how far you can push your logic models, without a way to put them to the test).

For a scientific theory, "being easier to grasp intuitively" may be the basis for more rapid advances over the previous version based entirely on a complex formal notation.

Ok I think we mostly agree, so… /is/ there anything in there that could be tested, even in principle? Or make any sort of prediction about anything observable? It all seemed quite vague to me…

You are looking at an idea which is pretty much starting to get defined as you watch it. Mathematicians are still trying to map all implications of this. Don't expect that everything is known already.

It's essentially a pitch for the Wolfram Physics Project. How useful is it to work through a system which has some probability $p$ of modelling the basic nature of the universe to some degree $d$? Reasonable people may reasonably differ on the answer depending on $p$ and $d$, and on the values of $p$ and $d$ in any given instance.