The downside of the design is that if you have a fault in the OP tube, or in fact turn the DAC on without OP tube
plugged in, full B+ will be across the cap - not a good thing!
Generally speaking a good design should have caps rated for full B+ operation or otherwise have some form
of protection for the cap.
Hi, I took a step back from this thread for a few days due to the direction it was heading, which was partly my fault, I apologise for any angst my posts may have caused.
The above point made by zenelectro is a very important, and further illustrates the need to choose a capacitor rating that is suitable for all possible scenarios of your circuits operation.
In addition to the solutions offered in this thread, and if you wanted to keep using the 200v rated duelunds, another option would be to reduce the B+ to below the capacitor rating, and use an active plate load/CCS (such as a cascoded IXYS 10M45S, DN2540N5, etc). This would avoid the need for a large plate load resistor (and thus the high B+ to allow for the drop across the resistor). This is very simple to do, basically 2 depletion mode mosfets, gate/grid stopper resistor, and a source resistor to set the ccs current, 4 parts per channel. And at the currents required for your circuit heatsinking would not be required. Another side benefit would be a reduction in output impedance for the stage.
edit: although I expect a ccs plate load, and/or the zener string/clamp, will both have a large effect on the sound/voicing of this dac that Steven et. al have worked so hard to achieve. But, still an interesting experiment to try - if the ccs load hasn't been tried before, it may have been.