Update! 8/22/99 Ceramic capacitors can, and do, explode!
I am also working on a capacitor of my own design, which should be able to handle 100,000 VDC. It will be a stacked plate capacitor, with forty-eight 6" by 14" plates made from aluminum roof flashing. The plates will be stacked with 12 inches of overlap and separated by 15 sheets of 0.006" linear low density polyethylene. The trick is to stack the plates and sheets in a pan (also made from the flashing) and then run the whole lasagna through a vacuum oven at 140 ºC. I built a small test capacitor using this technique, just to make sure it would work, once I realized how much effort it would require.
Next, I hooked it up to my 30,000 VDC, 3 mA power supply, a spark gap, a multimeter with 40,000 V probe and a 23 watt, 23 megaOhm resistor (to limit charging current) and tested it to destruction. Starting with the spark gap set for 10,000 V, I charged the capacitor until the gap fired and then increased the gap in 1,000 V steps. At 22,000 V the spark gap flash and bang was nearly as violent as the spark from my late 20 nF commercial cap at 30,000 V! (Can you say "low self inductance"? I knew you could.) Finally, at 23,000 V, a weak pop emanated from the capacitor and it would no longer hold a charge.
Other than trimming the corners, I took no special precautions making this test cap. For my big cap, I also filed the corners smooth and then polished the edges with a 220 grit sanding pad. The slowest part has been checking each plastic sheet for defects. I am using the sort of film normally used as a vapor barrier in housing construction and it has lots of defects, including holes, gels (little ball shaped particles of cross-linked plastic) black-specks (flakes of burned plastic and metal filings) thin spots and even a few pieces of steel wire. I hold each piece up to a light and reject any (about two thirds) with obvious problems. Some of the better rejects will go on the top or bottom of the capacitor in stacks of 80 sheets (~1/2"). With 15 sheets between each plate, instead of the three in my test cap, it should be able to handle 100,000 VDC with no problem. Whether or not it stands up to Tesla coil service remains to be seen.