This has always been a problem -- switching kiloamps at kilovolts and getting it done quick without a lot of series impedance and a lot of damage to the electrodes and other parts of the switch.
A normal spark gap has some real failings here. The way the current flows in a coaxial one tends to compress the arc to a tiny point, limiting in "bead" instability where the plasma winds up as a tiny bead, and on the way to that state, the current density on the electrodes becomes insane, blowing pits out of them. This coats the insulators with conductive junk, and messes up the electrode end shape very quickly, so the next shot isn't the same (or sometimes even possible). I've been looking into inverse pinch switches for improved performance, as I have some really decent kilojoule capacitors to play with here, and would like to do some pulsed work myself.
Note that if you're not making a switch per se, that this plasma compression (Z pinch) can work for you. When I was a pseudo grad student, we made a rig that would compress a plasma evaporated off one of the tungsten electrodes down to a single bead about 1u in size, a very nice point source of X rays at far higher energies (hundreds of kV) than the cap voltage (10kv). As far as I know, no one has tried this with a D fill gas for fusion, everyone does dense plasma focus instead, and I'm not sure why, as this would seem to give a much higher "compression ratio" and no question the energy is there...
Here's an interesting bit of earlier work on an inverse pinch switch. Later efforts used a more mushroom cap shaped center electrode and claim benefits from that. This paper seems to think that letting the arc expand out ruins the inductance of the switch. I am supposing we'll just have to test that ourselves, as this literature is "dual use" and most of it is hard to find. On top of that, it generally just describes one setup and the numbers associated with it, so it can be hard to do a decent comparison and do a good design from the literature alone -- you just have to dive in and try some things. I put this one first because doggone it - it's about the only one that has a really good mechanical drawing showing how these are made in general. In this and most other designs, breakdown starts across the insulator around the mushroom stalk, then moves outward as the magnetic field pushes it that way. Later designs let this proceed up the walls of the outer conductor (which isn't insulated) and around to the top of the mushroom cap before extinguishing -- which spreads the damage around better. In other words, one of these looks a lot like a DPF fusor inside out, and in some cases the discharge self terminates at some point. This can be a good thing if you want it, saving stored charge in the capacitor, but having it shut off at the right time for whatever you're driving becomes another design issue to have to handle.
Seems most of this NASA work doesn't allow for the discharge to go up all the way to the top of the cap -- they have their trigger electrodes in the way, and in general most of these are designed for higher voltages than most of us would use as well. Interesting anyway.
I have more of this someplace, but my library can be a little daunting to fish through. I'll add it when I find it in a later edit.