Hi Vladimir,
I sympathize with the idea that everything (including the vacuum, energy, matter, antimatter, radiation, dark energy and dark matter) should emerge from the interactions of simple units. We have very good examples of complex phenomena resembling particle trajectories or life-like patterns, emerging from interactions of elementary, binary cells arranged in 2-D (Conway) and even 1-D (Wolfram) arrays. Your Kepler packing of magnetic dipole spinning nodes reminds me of Fredkin's theories, e.g. his 3-D Reversible Universal Cellular Automaton model.
The hope of these approaches is to explain the highly complex in terms of the very simple, and I believe that the powerful notion of emergence (in computation) might indeed satisfy this need.
But what I find annoying in cellular automata models, as well as in yours is that you assume (if I understand correctly) an infinite, pre-existing, regular lattice, for the game to start. That's very costly a structure to set up in one shot. Wouldn't it be much cheaper to have the 'ether' structure be built progressively, starting from (almost) nothing? In my experiments with algorithmic causal sets, meant as instances of discrete spacetime, the structure is a directed, unlabeled graph that grows. Depending on the underlying algorithm, some regular, grid-like structure can develop, on top of which localized structures may start playing their interactions; both the 'background' and the actors are produced by the same mechanism.
Also, you attribute some state properties to your nodes, some labels, e.g. spin, in the same way as cellular automata assign a binary state (or ternary, in Fredkin's models) to the grid nodes. Let me just mention that, if one is concerned with (extreme) minimality, there is an alternative approach, that I explore in my essay here: one could try to use stateless, unlabeled nodes, and hope to see everything emerge from the dynamics of the graph growth. Ciao. Tommaso.