This is a general question, although the immediate purpose is to apply it to a Thorens AZ-25 amplifier from the 1960s. Schematic is here.
Push-pull valve (tube) amplifier output stages are generally either automatically self-biased via a cathode resistor or else have negative grid supplies which are adjustable.
The purpose of adjusting the bias is to set the standing current so as to get the stage just into class B, where the conduction angle of each valve is exactly 180 degrees, to minimize crossover distortion. Increasing the standing current any further takes you into class AB which is no further improvement, and a waste of standing current. (Some people e.g. Bob Cordell call the correct adjustment point Class AB, but in my view they are mistaken.)
In general this is pretty easy to do using a distortion analyzer: set the standing current to zero, or at least too low; feed in a sine wave; observe the output across a dummy load; and increase the standing current until the crossover spikes just disappear.
I've carried out this adjustment many many times on transistor amplifiers and it is easy when there is only one adjustment, typically a bias spreader.
This specific amplifier and others have separate adjustments: one for the upper valve's grid and one for the lower valve's grid. In this case I'm a bit lost for a definitive procedure. The idea is to ensure that the standing currents through each valve are equal, but achieving that seems like an endless back-and-forth procedure to me which may never converge.
My question is really is there a kind of analytic solution to this where you would only have to make two adjustments? or at least be assured of fairly rapid convergence on a setting?