HMMMMMM. Hmm.
Well I had to try the triode-strapped *tetrodes in the DAC tonight.
The big reveal! It's the 6E6P-DR.
Very linear when triode-strapped, mu of 33, plate resistance 1.2K.
Wired up another pair of tag boards, ferrite beads on all electrodes to discourage oscillation. Thought I'd do a whole new pair of tag boards in case the change isn't to my liking.
There it is. I/V resistor reduced for the same output voltage, batteries changed from 12V to 1.2V, CCS loads reduced from 30mA to 25mA for my chosen bias point.
So what does this do for us - well the THD came down significantly - with a -6dB undithered 1kHz sine (again), THD is 0.018% compared to 0.07% with the 12B4A.
I haven't measured it yet, but the output impedance will increase from ~750ohm to ~1.2K, but really not an issue.
The high frequency response, however, is more rolled off in this setup. This is likely due to the increased input capacitance of the 6E6P-DR relative to the 12B4A.
20kHz is down -2.7dB relative to midband compared to -0.9dB using the 12B4A.
Subjectively though, I do think the clarity and detail retrieval increased quite nicely. I really don't get the impression the top end is rolled off. Tonally I would say the change shifts the sound of the DAC more toward neutral if I were to call the 12B4A setup warm. In addition, the noise floor is lower and the 6E6P-DR is less microphonic than the 12B4A.
Despite the measurements, I would call it an improvement. I think I will leave it as is for now without making a final decision and listen for a few days. I still have the 12B4A tag boards, so could switch back for another listen, which I will likely do!