The keys are needed drive voltage and the amount of feedback to be used. If it's an A2 PP KT88 type, then about 142V-PP drive is needed at each plate. This also assumes ~525VDC at the plate, triode mode, ~65mA idle current (-61V bias) So, allowing for excursions to +10V on the grid (which is probably extreme), I get ~142VPP.
If you DC couple the pairs, then the second plate will be somewhere around 300V (cathode around 150), so the + swing will bring the plate to +371V. Better have at least 425V B+ with plate CCS/choke loads, or >500V B+ with resistor loads.
The other application is G2 screen drive. Now we need 300V PP drive to each grid
So that 300V becomes 450V+ at peaks. 500V with choke/CCS loads, or >650V with resistors!
If you assume 3V PP as an input, then the screen drive amp needs a gain of 100 from the input stage after feedback; the A2 amp needs 47.3 after feedback. Another assumption comes in here, that with a good set of output iron, either amp will only need 10dB of feedback at most. If that pans out, then the gains needed become 316 and 150 respectively. I think.
So, with those requirements, and assuming and optimized gain 2nd stage consisting of a 6SN7 LTP (it can swing the volts cleanly!), with a gain of close to 20, then first stage gains needed are 16 and 7.5.
Since the fisrt LTP isn't fed a balanced signal, it's gain will be roughly 1/2 of mu. The 6SN7 works with the A2 amp (x10), but the screen drive needs 32+ (6N7, I think?).
Now there are a lot of "ifs" in the above, and that may mean a 6SC7 or higher gain tube. Only building, testing and listening will tell...........................................................................
Slick