Another approach to a low-distortion AM demodulator was the combination of a diode direct-connected to a cathode follower.
It was mentioned briefly by Langford-Smith (page 1495 in the “Classic” edition), with more detailed coverage in AWV Radiotronics for 1953 June, pertinent page attached:
A more detailed analysis was provided in AWV Radiotronics 1956 January, page 03ff. This is available at:
https://tubedata.altanatubes.com.br/...ics/index.html.
This suggested that the reflex demodulator (another name for the infinite impedance type) did not offer much improvement over the diode, whereas the diode-plus-cathode follower combination did.
I imagine that for the diode-plus-cathode follower combination, one could use a 6BJ8, 6BN8 or similar double diode-triode with separate cathodes for each section. The second diode could then be used as the AGC rectifier with free choice of delay voltage. Perhaps an EABC80 or similar would work, with the independent diode used for demodulation. Whether one of the other diodes could then be used for AGC is debatable, given that its cathode would not then be grounded at AF signal frequencies. On the other hand, extra smoothing in the AGC circuit, as used when the AGC bias is taken from the signal diode, might work. AGC considerations aside, an ECC82 with one triode strapped as a diode would also be suitable.
Cheers,