It sounds very similar to severe underbias in the output stage. Being a valve driven MOSFET output stage I'd guess there is very little negative feedback to elliminate crossover distortion. Most amps bias generator networks are wired so if the bias pot goes open circuit (usually the case from contamination of the tracks) the output stage will loose all bias. MOSFETS needing several volts of Vgs will suffer from severe crossover distortion as there will be a considerable amount of signal around the 0v crossing area on the output that will not be amplified. There are alot of distortion mechanisms that come into play in such an instance but let's not worry about them hey
You should be able to check across the source resistors with no signal applied and no speaker connected for a couple of MV of voltage. If you measure 0v then this is the case. If you can identify the gate connection then measure the voltage difference between the gates of two FETs from opposing power rails, if you measure 0v again it is definetly the bias generator that needs attention/ re-adjustment.
You should be able to check across the source resistors with no signal applied and no speaker connected for a couple of MV of voltage. If you measure 0v then this is the case. If you can identify the gate connection then measure the voltage difference between the gates of two FETs from opposing power rails, if you measure 0v again it is definetly the bias generator that needs attention/ re-adjustment.