[Amps] Need advice on DCblockinig caps-My Text Body Didn't Come Through!

[John Farber]

 I have a 6 m amp, a HB model using a single 8877, a Drake L4 box modified by a guy "back east" that failed. Big arc in the RF deck, followed by a loud arc in the remote PSU. I replaced all the diodes, caps and equalizing resistors in the PSU with components at least as good as the existing components. A few of the old caps had rounded tops, and a few of the equalizing resistors were way off, and the diode bridge was toast. I saw no obvious arc paths or other telltale signs of problems in the RF deck. Powered up the amp, let it sit for perhaps 5 mins, let the 8877 3 min timer ckt kick in, etc.  Applied a small amount of drive, got output, looked good on the scope, so I knew the tube was still good. Slowly increased drive pwr, up to about 1 kW out, when bang! arc in the rf deck, followed quickly by an arc in the PSU, and I quickly shut it down. All this in about 15-20 seconds. Let it sit for a couple years, as I have had some medical problems. Decided to pull the cover off the RF deck today and have a look-see. Again, nothing obvious, BUT upon closer examination of the DC blocking caps, the builder used two small doorknob caps in parallel, with an American brand name, 1000 pf 5 kV each. So I figure the two parallel caps is a good design, splitting RF current, and 1000 pf is about the right value. But It seems to me that the 5 kV rating is way too small. This amp runs 4 kV of B+. I seem to remember that the tank ckt can reach X 2 times the B+ value, which in this case would be 8 kV and way beyond the max rating for these little doorknobs. So I'm asking for advice on selecting suitable replacement caps. I found a site selling 970 pf, 15 or 20 kV Ruskie doorknobs for a not-too-bad price, and I remember hearing folks remarking they were happy with the Rusky doorknobs. What do the esteemed elders of this Amp group recommend? And what other components should I be concerned with to beon the safe side when rebuilding this amp? My plan was to examine the rectifier/filer board for failed components and fix that. Then disconnect the  B+ line to the tube and test the B+ by itself for an hour or more. If good, reconnect it and see what happens, and if all is quiet, apply rf and test for output. BUT that's what I did the first repair attempt, only to have things blowup, and it cost me about $100 in parts. So I figured I would ask for advice this time first.
Thanks, John Farber, KG6I