The HLMP-6000 was selected some years agon by John "Buddha" Camille, for its exceptionally low noise and its low dynamic impedance. Other diodes will have different voltages and impedance - by the way, there are several different red formulations.
The direct couple circuit has some negative DC feedback to stabilize the operating point, derived by feeding the EF86 screen from the 6922 cathode. (Incidentally, the 'Tode guitar amp under development has the same circuit - an old Mullard innovation.) Most old-stock EF86s are close enough to the spec that they will bias up correctly with the design choice HLMP-6000. Modern production EF86s are less consistent and many need a greater amount of DC negative feedback, provided by the servo circuit.
Obviously tubes that differ further from the EF86 specs can sometimes be made to work by the servo, but would be especially unlikely to work without it.
An alternative to the servo bias is adjustable bias, in the present case substituting other diodes until the right operating point is achieved. Unless you have access to tube curves at 100v on the screen, it is impractical to predict what the bias should be - I had to develop my own EF86 curves to do the original design. I have another adjustable bias in mind, but have not had a chance to test it yet - and I won't share it until I know whether it works, sounds good, and is quiet. Risks of changing the bias scheme are noise and distortion. Risks of changing tubes are noise and RIAA equalization accuracy - the tube's plate resistance is incorporated into the equalization, primarily affecting the 50Hz low frequency corner.
We elected not to pursue such a design because it requires re-adjustment as the tube ages and whenever tubes are replaced, including rolling tubes. It's not just the practical difficulties, I also worry that too many Eros' will be operating out of adjustment, sound bad, and create a poor opinion of the Eros.
