Just got a Vox Supreme from the original owner

[R.G.]

Good possibility.
The Thomas Beatle amps had a similar issue - overheating, low air flow, and small heatsink area. IIRC the Supreme uses a series pair of output transistors in both positive and negative sides to deal with the high voltages from the power supply, where Thomas went parallel at lower voltages. The heating issues in Thomas Beatles vanish when you parallel each output with another output device. This cuts the dissipation per device in half while leaving the heatsink to air path untouched. But that reduces the temp rise from chip to heat sink by 10-12C and nicely cuts chip temp.

[Mick Bailey]

I just had an interesting conversation with the original owner about the distortion volume drop - this is normal for this amp from new (which we've previously established from the schematic), but there's more to it: When he got the amp he was using it with a stereo Gibson 345, split between the clean channel and second overdrive channel and effectively channel switched by the guitar. With both channels set to the same clean volume, when switched to overdrive, despite the initial louder attack of the clean channel, the overdrive channel sustained for longer at a more constant volume and this gave the impression of overall equal volume for the listener. Also something I'd overlooked is the 4x12 cabinet had two Midax H.F. horns which emphasised the distortion harmonics, and I've been using the amp through regular speakers.

[loudthud]

R.G. What do you recommend as replacements for the output transistors ? I don't think commonly available 2N3055s are up to the task.

[R.G.]

R.G. What do you recommend as replacements for the output transistors ? I don't think commonly available 2N3055s are up to the task.

They're not up to the task. I would use one of the MJ15xxx series. The MJ15022G or MJ15024G should be great for this. The MJ15003 has similar beta falloff at high currents to the 3055s - which is not a good thing, particularly. All of these have much bigger Vbceo and SOA than the 3055. Those are direct pop-ins for the original TO3 packages. If you allow going for TO3P packages, you can get into the Toshiba and Onsemi power-monsters at about 30% less money per device.

The 3055 just doesn't have enough HFE linearity at high currents for this app. and it doesn't' have enough BVceo and SOA to use only one, not two in series. It's possible that the higher voltage MJs and other TO3Ps would be able to eliminate the stacking. But for even higher reliability, paralleling more devices will dramatically cut junction temps on the outputs.