I'm not sure and would guess that such a system might be unstable in the long run.
You know, I was thinking that would turn out to be the case myself. But I started playing with a gravity simulator I've got here. I dug up a preset module simulating of the orbits of Epimetheus and Janus, thinking it'd be a quick fix to just alter the mass of Saturn and the two moons (and their orbits), for a fast and dirty test run.
Just by accident, when I found the Saturn presets, I also stumbled across a series of simulations someone else had run. He was just introducing a cloud of a few hundred "moonlets" equal in total mass to our moon, into orbit around a moonless Earth. It turns out that occasionally the debris cloud would condense into two moons of equal mass, each half the size of our moon, and in stable orbits.
*Well, let's say
close to a 50/50 split in mass. Some of the debris got thrown into space, and some hit Earth while the bulk of the cloud was jockeying for position. But the bulk of it ended up in the two remaining satellites.
When this did happen, it tended to consistently be one with a tighter, near circular orbit, and the other with a much wider highly elliptical orbit. He never got anything resembling the freaky relay race style orbits of Epimetheus and Janus, though.
If my insomnia is still in full effect tonight, I'll play around with it for a few hours and see what I come up with. So far though, it doesn't look nearly as implausible as I was expecting.