|The drawing above shows R26 and C10 in red as adjustable components. R26 and C10 were not present in the first Williamson diagrams. They were added when oscillation at high frequencies was noticed as widespread a problem. The right values of these components depend on a lot of circumstances like the tubes that are used and the exact electrical values of the output transformer.
One can expect that values mentioned in the Williamson diagram were determined with the 6J5 tubes, the KT66 and the output transformer that was used to determine these values. The components do not add a special quality to the circuit but solve a problem by spreading out the frequencies where the phase shift may turn the negative feedback into positive feedback. In general this happens at frequencies around 100kHz.. In general the amplification factor of Williamson starts to increase from 80kHz on.
When the output transformer and the electronic circuit with the tubes have the same frequency where phase shift reaches maximum oscillatiomnmay be inevitable. This effect can be boosted by the fact that the loudspeaker impedance goes towards infinite at frequencies around 100kHz.
To determine the right values in your case (with your tubes and output transformer) you can take a course in theoretical physics and calculate forever or solder a potentiometer and a variable capacitor (from an old radio) provisionally into the circuit. connect a signal generator and an oscilloscope (with a dummy load) to your amplifier and turn the knobs of R26 and C10 while watching a square wave of 10kHz on your scope.
Your goal is to avoid a large overshoot of the square wave and a clear recognizable sine wave form. Go for some fuzzy waves on top of the squares. Try to force the thing into oscillation, then you clearly see what you have to avoid. Once you reach the situation with little overshoot and some fuzzy waves on top of the square wave determine the values of the potentiometer and the variable capacitor and solder the components with these values in your circuit.
With The Ultra Fine output transformers the best situation was when to whole R26 and C10 circuit was left out. See the image below on the right, there is not much to improve!
With the Unitran output transformers the best result occurred with R26 was 22Kohm and C10 was 220pF. The image below on the right is with the values Williamson prescribed. If you count the waves on top of the squares you can predict that if oscillation occurs it will be at approximately 80 kHz.