This http://www.untrue.ch/test/ehx_12ay7/ehx12ay7.html is as close as I can find to any type of photo representation of what is inside. You can see the two transistors on the shared heat sink in the top toward the right hand side of the first photo. Plus an IC in the upper left. Who know what EH uses; if they have a house number or a real world number. If you had the right test equipment you could find out what was wrong, or a schematic could let you signal trace it.
If the IC has a real number on it, Google the pin out, and ‘scope the input and output pins if you had a signal generator.
It almost looks like a plain TLO74. Wow. I wouldn’t expect such openness from EHX, but this is a pleasant surprise if it is true. Here http://circuitdiagram-schematic.com/12/pre-amp-equalizer-using-tl074/ is a preamp schematic (NOT the EHX, but just one using the TLO74) to get the pin out and an idea of a typical circuit.
That’s all I know. With real part numbers, you can test anything, and know where else to look. I have had this happen before, and just plain dumb luck helped me. And this was with supposed name brands that had a ‘reputation’. If you have nothing to lose, go for it!

I said ‘close to it’, and there’s no reason for me to back down. It can be done. I’ve done the 6V6 in a Twin trick, and it works. Tubes have a different impedance, but most of the tubes we use today are close to being neighbors that it doesn’t enter into my mind. And you can load 6L6’s at almost any impedance you want. The old stuff was up to 10K and Fender went down to 4200-ohms. I’ve bought 3800-ohm OPT’s and they work just fine. The power output goes up, and so does the bandwidth and distortion. I’ve measured 6V6 OPT’s anywhere from 6600-ohms down to 5K. I built a 4 x 6V6 amp and used a Twin OPT (which is where I got the 6V6 in a real Twin idea) and it worked. That’s 2100-ohms. Marshall used 1900-ohms. That is too low, even for a fool like me. But anything else? Fair game if you crank the Plate current down to nothing and check everything first before adding heat.
No, this is where I have more test equipment than God, less common sense than a soap dish, and the experimental curiosity to be dangerous. If you have sense, maybe you aren’t as foolhardy as me, but what you can learn in one evening is worth more than spending a year in front of a computer reading Forum posts. The GEC KT88 data shows 4K for a pair, 460VDC, and generating 100-watts. So maybe, and that’s a big maybe, the power transformer can’t handle the extra filament current, or the Plate current. Biased up properly (no crossover notch), maybe the output transformer can’t……..
Ah, just measure everything, and maybe you can’t do it. But I wouldn’t say ‘NO!’ until I measured everything. That’s what I do anyway, just to kill a lazy Sunday. Remember many Marshalls came without any Screen Grid resistor at all, and they’ve lived this long. It’s up to tube quality, and how hot you run them. If you measure the voltage drop across a Screen grid resistor, you’ll know if it’s up to the task. Remember, the smaller the value, the less of an IR drop, so lower wattage can be fine.
Without knowing what Mr. Grim knows, I shouldn’t kept my mouth shut, or at least tempered the answer. But still, if I put the bug in his ear on learning more, then my work here is done.

Just a SWAG; it’ll either be OK or possibly run a little ‘cool’. But I’d get a simple little bias probe type tool and check, and be safe now rather than sorry later. I believe you can run a little more heat through a 6550/KT88, so it should be OK. It’ll have a lot more headroom and probably more bottom end, so let’s hope that’s what you’re after.

Then my only other SWAG has something to do with his harmonica playing. But that’s it; I give up after that poor guess.

If it’s an input capacitor discharging as the footswitch connects the input signal to the first stage (as in a true bypass scenario), a pull-down resistor soldered between the input circuit and ground gives the capacitor a path to gently charge and discharge. Any high value will do, but something like 2.2Meg or 4.7Meg seems popular if you haunt other pedal forums.
Also, that second link tells you the LED charging may also contribute noise, so the ‘mod’ there handles that with a current limiting resistor added and an extra filter to smooth things out.

I can’t speak for the ‘U’ version, but the ‘H’ is just a JCM800. Very simple circuit. I doubt The ‘U’ is that drastically different; likely the voltage selector is the one and only difference. It might be something really simple, and I’d treat it as a JCM800 and take the ‘send’ from the Cathode follower. If your ‘tech’ doesn’t want to do the work, he should just say so. But to make you find a schematic as a cop out is being lazy. If I’m reading this wrong, I apologize now, but I still insist it’s likely a simple circuit, so just do it already.

You’ve added true bypass switching, but did you do anything else? It looks like you have the collector on the ‘left’ lead of the transistor, and I’m not sure I’ve seen many transistors that do this. It would appear the pin configuration you have is c-b-e, and it’s usually e-b-c. Everything else looks OK, unless the switch is wired wrong.
Does your transistor say 2N5058? It’s difficult for me to read that, but any 5058 I’ve seen is a TO-39 or soup can.