I like the sound of "something strange going on." Heck, it might be worthwhile to try various mixtures of KCl and NaCl to see if there is an obvious particular ratio where the sparking is maximized. Mills uses a pinhole technique for finding UV where he actually drills a sub-mm hole into the reactor wall with line of sight to the plasma, and mounts and glues a photocell UV detector chip over the hole - but that is for a vacuum reactor. Robin may have experience with this...?
Michael Foster wrote: Some of the sparking looks a little too far from the surface of the glass container to be contaminated by it and there's a lot of sodium spectrum being displayed. So something strange is going on, I think.