printed circuit caps

To return to an old story. Some weeks ago I started a discussion on using
fibreglass PCB material for pulse caps. A number of people expressed doubts,
but I went ahead anyway. The second cap I built was moree or less the same as
the previous one, but this time boxed and better insulated. Value 2n2F
(measured). I ran it at about 25kV for a total of ten hours before fitting it
to a small coil (2.1" X 15", 800turns, 4.5X1.3 toroid) Big mistake was that for
speed I used a cylindrical primary ( wasn't expecting it to work too well).
Result:- consistent 3.5 - 4" arcs, but I cant't open the static gap fully as
there is arcing between the primary and secondary, low down. By the way the
driver is solid state using a LOPT, in other words about 20kV -at-1mA.