I've had a little time today to think more about the regulator-in-the-cathode proposal, and after further thought, I've come to the conclusion that neither a voltage regulator (ala EFB(tm)) NOR a simple current regulator scheme achieves the nirvana that folks are looking for. In fact, the simple resistor comes out looking pretty darn good against both of them. Consider:
1. Using EFB, there would in fact be operating point compensation for differing B+ levels (AC line voltage fluctuations) -- just as a simple cathode resistor will do -- but the initial bias would still have to be manually set for the correct optimum quiescent current flow at a B+ of 265 vdc, FOR EACH TUBE INSTALLED -- just as you need to do with the simple resistive bias controls now shown on the schematic. Therefore, this approach isn't really gaining anything over the adjustable resistive approach now used.
2. A CC cathode regulator will surely keep any and all tubes dialed into the correct optimum quiescent current flow -- no matter what the B+ level does, which is a problem as well. If the B+ drops due to a lower line voltage, the CC regulator will react to keep the current flow through the tube the same as before -- except that with a lower B+ level, maintaining the same quiescent current as specified for the normal B+ level distorts the load line, since the quiescent current can't fall in proportion with the reduced B+ voltage -- which a cathode resistor WILL allow to happen. If the B+ rises above specification, the regulator will again keep the current draw at that required for the specified B+ level, distorting the load line the other way. So using a CC regulator presents its problems as well.
To achieve the nirvana that a CC cathode regulator promises, the B+ to the stage must be regulated as well as its quiescent current. With that voltage held constant at the specified value, and the CC regulator dialed into the correct optimum quiescent current flow, then you can in fact change tubes to your hearts content, and they will always be perfectly biased -- each and every one of them. But understand that this is the ONLY thing this scheme can do (compensate for individual tube characteristics). It can't provide any better performance over that of a properly adjusted cathode resistor, and as discussed earlier, it can't do a very good job of compensating for a circuit that's not designed for true Class A operation to begin with, either. In the end, optimizing this approach starts to get pretty far away from the original concept that the little 8600 represents real fast. Against that, the simple adjustable resistive bias network shown on the updated schematic looks pretty darn attractive -- and particularly so because it can produce just as good of performance as any of the other bias schemes can.
Dave