I can't see any reason why using parallel feeds (rails) for different stages can not be used? Perhaps there comes something positive when typical series feed is used and stages interact? But if there comes interaction I think it is mostly between B+1 Output Transformer/Anode(s) and B+2 Screen(s). Because for the Phase inverter B+3 filter feed there is usually 10k dropper and the capacitance is quite high for much smaller current. And following stages B+ usually at least 10k.Here's another idea. I have an AS-05T240 in my junk drawer. It has both a 240V and 220V HV tap. I can run the 240V tap in the amp and beat the crap out of the EF80 plates at ~300V. I think Rob has stated before that he's run them at 300V. This will also bring up the rest of the voltages at the B+ nodes. I could also use the 220V tap and raise voltages marginally.
So after first additional filter capacitor (which now becomes a B+0) divide feeds to two "rails" one for B+1/B+2 and another for B+3, B+4... Then B+3 voltage comes higher and then B+1 can be dropped lower using a Sag-resistor.
Now the Sag-resistor and B+1 filter capacitors can be "tuned" how you like. Perhaps a potentiometer can be good for testing if powerful enough can be found? Or just a SPDT switch which has an Off in the middle and three different value resistors?
Perhaps instaling one more tube socket for 12AU7 does not make EF80 amp too complicated.
Now I have been thinking more... if you have a CT power transformer it is possible to get solid bias circuit and using fixed bias for 12AU7 make it sound even better. And if you wan't both require only one SPDT switch to change between Cathode and Grid bias.