I found some papers on electrostatic cone-trap mirrors that got me a bit fascinated with the idea of beam on beam fusion that solves some of the thermalization and other problems ChrisMcB reminds us of. No one has done fusion with either yet, but it seems like this has some potential with appropriate additions. What's neat about electrostatic-only things such as are described here is that you can have counter-motion beams, particularly in the "loop" configuration. One could add bunching/re-acceleration timed correctly at a point (or two) on the loop to keep things moving and in focus so things that missed would get more chances at recirculation. Even though the beam densities are on the low side here, it looks like it's worth playing with a least a little, as the numbers on how many passes ions can make before being lost to collisions (when they set up to avoid those!) are pretty darn high. Of course, in our case we want collisions, ion on ion, they are mostly worried with storage times and neutrals scattering things out of the beam(s) in these papers. I found a reamer at McMaster that "just happens" to cut the angles they are using in these papers, which I kind of doubt was complete co-incidence, and am starting to make some up for trial in my big tank...why guess when you can know?
This first paper describes the simplest possible system, maybe not so useful, but it makes the basic understanding needed clear.
Once you understand the first one, this is what you'd really want to build, but add a couple of bunchers to keep the particle speeds up and refocus them as they go around. Timing would of course be important, but at least with this design you don't need two rings to collide same-charged particles.
More when I know more -- I'm still doing parameter sweeps on fusors at the moment and working on some really serious data acquisition and analysis tools which will be good
no matter the final approach I settle on. Ideas are always far easier to have than real working things