Phono is not really "high gain". With the simple NFB EQ, if your midband gain is 30dB, the 20KHz gain is 10dB and the unity-gain point is 3X higher, 60KHz. That gives a big fraction of a db "error" at 20KHz.
At 40db midband gain, some hot cuts will give quite large level out of the preamp. In days of slewy chips, I liked simple discrete. Smarter/richer designers in Japan favored quite high rail voltages to allow high gain (for accurate EQ) without stress.
Way too many ways to skin this cat.
Phono (and tape) are unique in audio because of the HUGE difference in gain from bass to treble. Like a tone control turned to both extremes.
Again, "mike preamps" are not a good fit. Mike outputs are 150 ohms SO-THAT they can drive long cables with low losses. In transformerless design, that leads to a preamp with low voltage noise and high current noise. But phono carts are 500 ohms in bass (so what?) rising to 10K-47K in treble (hisss). You want a preamp optimized for lowest hiss in 10K-20K sources. As a rough-hack, reduce input device current by a factor of 100! However this probably screws other aspects of the design.
> small buffer ... make capacitance a non-issue
Recall that most needles "need" some specific capacitance load. The half-Henry inductance would cause a fall at the top of the audio band, even with no C. And C is/was inevitable. So the 47K ~~300pFd loading bumps the last half-octave, AND seems to also be selected to offset mechanical losses in the top octave. Snazzy preamps have selectable input capacitance.
> cable capacitance could be taken care of with active EQ if necessary.
Too darn hard, and shrill hiss. If you had say 4X the proper C, the top-resonance would be 2X lower, 7KHz-10KHz, and fall after that at 12db/octave. The very sharp treble-boost needed to "get flat again" is awkward and inelegant when simple short-cable techniques (or buffers) are possible. Leave the 2-pole top-boost to telephone long-line problems.
> post the RIAA encode circuit
Not the whole story. The power amp has some HF roll-off. The cutter chisel shank has a resonance. Cutter coil inductance may or may not limit actual coil motion. None of this can be allowed to "rise to infinity" or stray scratch/RF would overload the system.
I've worked with cheap lathes where 15KHz was hit-or-miss and 25KHz wouldn't happen. I'd seen somewhere a 50KHz limit on some Big Name cutter.
The 200KHz encodings are fascinating but always very-special and _were_ left-behind in the 20th century.
The VG66 plan shows (as JR says) a 12db/oct hi-cut about 32KHz. Maybe two? R25 R26, and R6 R8?
I also know that good (and bad) cutting engineers will fiddle the system for a good (or bad) final result. Disk-cutting is not some perfect translation. At some point it breaks-up into "impressionistic paint-tossing". A good engineer gives a good impression. But he needs a "good" playback chain to guide his fiddling. Thus part-db "errors" hardly matter.
> using a ua709 opamp in a critical audio path looks like there should be some room for improvement there.
Don't touch it. (as you say) They die easy. They have become hard to replace. And a well-used '709 is an EXCELLENT amplifier, as long as you don't need short-proof or lowest-hiss.
In fact I'd argue that few "improved" opamps can match the '709 for high-end boost. This plan seems to give gain of 40dB at the top of the audio band, or 2MHz GBP, or 20MHz for just-good accuracy. De-compensated '709 will do that. No classic GP opamp would. Some newer stuff would whoop its butt on paper, but considerable re-design effort for possibly no audible difference. (Hmmm.... some CFB chips "might" drop-in with some pinout differences... R16 would compensate the loop. Maybe.)
