Ok, I put it in (schematics attached). I used 2SK209 models for the J-Fet's and BC550/560C models for the BJT's.
The J-Fet's have a bit higher Gm than the originals, the BJT's have lower Beta than the originals so I think the differences will mostly wash out.
Result for open loop gain:
View attachment 110388
The original circuit offers 94dB open loop gain with appx. 1kHz -3dB point (including Mr Miller). Uncompensated we have around 10kHz -3dB. going to 600 Ohm loading drops open loop gain a few dB, nothing major.
Closed loop gain for the M-15B (which I used as it has more details in the service manual) ranges from ~ 6dB to 56dB, so NFB ranges from 38dB minimum to 88dB at minimum gain.
My suggestion of Mosfet VAS with bootstrap load matches the original above 3kHz and increases LF gain (@ 100Hz to 114dB or 20dB more than original).
With a Darlington VAS + additional input stage degeneration (100R per J-Fet) overall gain is lower but LF gain is boosted.
With gain set to 12.7dB (with input transformer this gives ~ 33dB Gain) the closed loop frequency response is identical. Unless stated this setting (gain control @ 50%)
With 300mV output (nominal signal level according to Manual) and 10kHz all three circuits simulate as ~0.0003% THD with H2 being the only observable harmonic above the "noisefloor" of the sim and the original circuit being marginally superior to both modifications.
Raising signal by 20dB (still 10kHz) gives the original circuit 0.0017% THD with H2 & H3 notable.
The Mosfet version (e.g. mine) has 0.0022% THD and the lowest H3 (~3dB less than Darlington/degenerated and > 6dB lower than original), with Darlington inbetween. Going to 600 Ohm loading does degrade THD somewhat, around 3dB or so.
Simulations at 1kHz or 100Hz would likely favour the modified designs, but the value of these advantages may be questionable.
For noise, the darlington VAS is marginally worse than the other two, nothing I'd loose sleep over. The observed 4nV|/Hz @ 1kHz Ein are not that great, however we need to remember that the 1:10 step-up transformer on the input would lower this tenfold.
Unweighted SNR at maximum gain and 0.3V output level looks -60dB. Note that this is at 56dB Gain (actually 76dB total gain including transformer), so that is a helluvalot gain!.
Realistically, the original circuits offers fine performance that gives up very little to modern monolithic Op-Amp's in the way it is used, despite the very simple and unfashionable design. Non of the suggested "improvements", mine included appear to offer any material improvement.
As is, in the original context it is a great design showcasing that "appropriate technology" designs do not require fashion complexification.
Mind you, in this mixing desk it drives a 47kOhm load (manual value) not 600 Ohm. To turn it into a stand alone "general purpose" Mic-Pre would likely mean to either add additional active output circuit or making changes.
But that is not what this circuit was designed to do, so it's unreasonable to have such expectations.
For a modern design, the discrete OPA may be substituted by a number of different monolithic Op-Amp's, I think the TI OPA828 would make a great substitute. The transformers turn up occasionally in on-line sales and likely are most of what makes the "sound" of the circuit.
Using 2SK2145 for J-Fet & BC850C/BC860C for a modern discrete version using the original design might be fun. Don't forget to avoid loading the output too much, use a buffer to drive 600 Ohm.
Thor
I feel slightly sorry for Bill Whitlock's parents, who I imagine must have thought that William Whitlock had a nice and memorable ring to it, and then he goes and calls himself Bill. 
