[Amps] Need some help

[w2cqm at juno.com]

I'm having a problem with an amplifier project that appears to defy a solution. Hoping someone can suggest an answer that apparently I'm not seeing. Except  for a rewound HV plate choke, the amp is a basic Drake L4B with a pair of 3-500's. It's been mated to a custom built  outboard Hipersil, step start, metered supply making about 4.2KV plate voltage. RF deck primary voltage circuit changes were made (done this before successfully  with several Drake 8877 retrofits) to control the power supply with the reconfigured  RF deck on/off switch controlling 110VAC.  On this retrofit, the grids were grounded and the filament voltage CT biased through a zener to the  B-. Pretty straight forward modification. On the initial start up,  half power HV was applied  via a Variac  and the (no signal) idle current tested. All was well and  indicated an appropriate idle current simultaneously on both the deck and power supply plate current meters. I slowly increased the plate voltage to full output. After a short time at this power level, there was a "bang" in and around the HV choke/blocking cap area. As a result of the arc over, no fuses let go but the zener was replaced along with a small choke in the bias line. To prevent a recurrence of shorting, the under rated OEM Drake blocking capacitor component package was replaced with a more substantial 500pFd/20KV unit and  repositioned with greater clearances to ground.  The flash over was eliminated but other problems developed from this experience. After making all the circuit changes, testing began again and at Variac controlled half power levels,  the Eimac tubes began glowing red with no excitation. The remote power supply has its own 220VAC plug so there's no interaction with the RF deck (which has its own 110VAC plug). Oddly, there is no plate current meter indication as the plates glowed!  With the primary Variac controlled voltage quickly shut down  the  voltage on the plate meters were immediately reduced to zero. In other words,  The plate meter(s) in tandem followed the removal of primary voltage clearly indicating a current draw on the DC supply. To diagnose the strange occurrence I rechecked all the wiring (keeping in mind that it was working OK initially).I replaced tubes with a brand new set;installed new zener(s); checked/removed  all by pass caps, ensured that the bias relay  contact  set was not welded,checked the parasitic choke assembly, etc. etc. Subsequently, I even disconnected the filament CT from the zener with no result. In desperation, I removed the plate current/HV meters from the circuit with no change.Currently, the only way I can get the HV to bleed off normally is to disconnect the HV from the tubes' plate caps.  Obviously, something is drawing current through the tube that  continues to elude me. Before I give up and lose forever  the "glow in the dark" tubes (in favor of an 8877  retrofit), perhaps someone can see the error of my ways and save the day. Any help would be appreciated. Ron W2CQM/3