Internet lore, repeated by respected colleagues, has it that when renovating Quad IIs and replacing the coupling capacitors to the KT66 grids, an additional 33pF capacitor should be added from the anode of each EF86 to its grid. This local negative feedback around these valves is intended to compensate for the loss of about 30pF leakage capacitance to ground in the original coupling capacitors, which, it is alleged, was designed into the circuit.
While the leakage capacitance was certainly present in the original coupling capacitors, the only evidence I am aware of to corroborate the claim that it was designed into the circuit, and needs replicating, is that Quad Hi-Fi have been known to do this themselves when servicing.
Against it consider these facts:
1. There is already about 30pF grid to cathode capacitance (Cgk) in any KT66.
2. If the intent is to reproduce the HF pole effect of the original leakage capacitance, it is only necessary to add a compensating capacitor to the first EF86. The output of that EF86 goes through the second EF86 with that pole already in effect, so adding a second pole to the 'bottom half' of the amplifier can only cause 2nd harmonic distortion at the affected frequencies.
3. The effect of the original leakage capacitance shunting the 680K grid leak resistors was a pole at 7.8KHz. This is extremely low as a turnover frequency for a compensating capacitor. The effect of 33pF from anode to grid of an EF86 is a pole at about 27KHz. The bandwidth of a Quad II is about 55KHz, dictated mostly by the Cgk of the KT66s and the output transformer.
4. Whether or not the original leakage capacitance was 'designed into' the circuit, or more likely tolerated as a necessary evil, the amplifier is unconditionally stable without it, and without either of these compensating capacitors.
So we don't fit them.