Been wondering if I should even bother posting this, but what the heck.
FWIW, I think this goes against pmillett's design philosophy as I don't think any of his amps utilize this. If you're a pmillett purist or other type of purist, then no need to read this at all, and just skip over it.
I would do something different. I would not do the signal attenuation at all...i.e., resistor in series wired to a switch. That's attenuating the source signal, and with a 15x gain stage, it's not ideal and rather silly, IMO. The ideal would be to reduce the gain by changing tubes. Since that's not really an option (I don't want to scrounge the internet looking for a tube that'll fit and work), I would use negative feedback.
Remove all that signal attenuation stuff, and build in adjustable NFB to reduce the gain, and should lower the distortion as well. Some people hate that, but I think some NFB is good. Again, pick your poison.
I haven't measured or done this, but it's a starting point.
I'm too lazy to draw a diagram for the entire SSMH so I'll describe with words, first, bypass the cathode resistor (R5 and R11) with a 1000uF/16V cap I think even a 10V cap should work, not sure entirely, but 16V is fine. Without R5 and R11 bypassed you will get signal degeneration, so the gain drops some in the default circuit. Bypass it, to avoid some feedback, and the circuit gain should increase. That's right, increase it.
Put in a 10K grid stopper, that is a 10K resistor between the pot and the tube, but put it right on the tube pins (5 & 6). Also, put another 10K resistor right on pins 5 & 6. Connect this to a pot, and then connect that to the output of the circuit, after the output cap and the 2K resistor to ground.
The higher the value of the resistance, the less NFB. So, wire the pot in reverse, basically. Shunt the wiper to the input pin and wire the amp output to the input pin. Take the "ground" pin as the output, wire so it connects to the second 10K resistor.
You can eliminate the adjustable NFB pot, and use a single resistor, once you figure out what you need. I would do some measurements and look for something that gives you about 2:1 for IEMs, or 2x gain. IEMs, ideally, should just use a buffer. Even an IPOD has enough voltage swing for an IEM that can put out 120dB/V. high sensitivity low-Z headphones, I would go for 2x to 4x gain. Higher end Senn dynamics, I can live with 6x to 8x gain. Measure with a 0dbFS file on your computer to a DAC that you know gives 2Vrms output. Feed into the amp, and measure the RMS voltage at the amp output if you really want. Another way is to put a scope on it, but most people don't have one. Measure the pot, add up the resistances and put in the resistor value you want, or leave the NFB adjustable as it'll make the amp more usable for various headphones.
Again, pick your poison. Some people hate NFB, but I think the extreme amount of series resistance coupled with a 15x gain to utilize an IEM or a high sensitivity low-Z headphone is silly.
YMMV.
I hope you find this useful, IMO, it makes the amp more useful for low-Z. With the 150mA Class-A bias, I think this is a good mod for the circuit. You don't need 150mA Class-A bias for high-Z.
Again, I haven't measured this, but that's what I would do. In theory it should work, but the resistor values may need tweaking. That would require sims, and testing, and measurement, something I'm not going to bother doing.
You can build in a switch to drop the NFB completely. Break the circuit and change the grid stopper to a lower value, and remove the bypass cap. Since this is right on the tube socket, you can put the switch nearby.