The stability issues for size in multiplayer will not be in the stress on any one system, but the constant exchange of information between the two systems. Keeping track of what ships are where, who sees what, and every play action's consequences takes seriously more and more synchronization between systems, and provides serious opportunities for desync and instability. It gets exponentially worse with each player added. They are probably going to get that network code down really solid before they dare open up the larger maps. At least, I would be extremely cautious about taking a step like that and risking so much customer discontent when it performs even more erratically.
I only worked on the actual networking systems involved while taking part in various large projects, but synchronous exchange of information is its own swampful of ugly challenges for serious network work, let alone gaming. But then again, in my mind, it has almost always been gaming that pushes forward little bits of the envelope of computer science and capabilities. From what I can see, multiplayer for complex games will follow some of the same progress patterns as general conferencing technologies, both corporate and consumer driven. Having watched what is and can be pushed through my various network hardware setups, I am expecting interesting results. The software hurdles are still more messy than anyone wants to admit, but there are hordes of clever folk out there working on that very thing.
