Everything seemed OK, so the next time it wasn't raining, I took it outdoors.
Although it was outdoors on a damp night, and the RF ground was a bit suspect, it kicked out streamers 4 feet long. This was using something like 2kW of DC power.
The streamers were incredibly hot and bright, but crooked, and flailed around a lot. A different breakout point might help with this. Without a breakout point, there were no sparks at all.
The charging voltage wouldn't go as high as I wanted it to. I couldn't get up to the full 1kV, the crowbar circuit kicked in around 850V. It actually triggered in the exposure above. This repeated short circuiting eventually burnt my power supply out - again :-( It seemed that the crowbar was set a bit low, and also the sense wire to the crowbar was getting quite a lot of EMF induced from the primary.
As for the power supply, it was just a pants design, I will have to redesign the short-circuit protection module, or just take it out altogether. The rest of the system has survived a few crowbar firings at full voltage when the protection module was burnt out and totally inoperative, so I think I might get away without it.
I've done a fair bit more work towards fixing these problems, and it's more or less ready to go again, hopefully at 100% power this time. I'm determined to beat the SSTC world record ;-)