>These instruments were mostly of academic interest – I can’t think of any band that used one on stage (but I could be wrong).

Very wrong. The Prophet 5, Oberheim 4/8 voice, Roland Jupiter 8, MemoryMoog, and especially the Yamaha CS80 were all available before the DX7, and very widely used on stage and in the studio.

But they really were stupendously expensive - far too expensive for mere mortals.

(There were also fully polyphonic string synthesizers, which had a very, very simple synthesizer under each note, but they were based more on vintage organ technology.)

So by the time the DX7 appeared, Roland and Korg had already produced semi-analog) instruments with digitally controlled oscillators and analog everything else.

The DX7 stood out because the keyboard responded to velocity, and it had an incredible 16 notes of polyphony. So you could play your fancy jazz chords on the epiano sound with the sustain pedal down, and the notes would respond to your fingers and wouldn't cut off too obviously.

The maximum for a pure analog polysynth was 8 notes, and cheaper models had between four and six notes. No fancy jazz chords. Especially not with sustain.

And the DX7 sound cut through - it was much thinner and more acoustic-sounding. The big analog polysynths sounded much bigger - lush, but less versatile, because they were best suited to string and brass fill-in pads, occasional special FX, and perhaps the occasional lead.

Yamaha did try to take synthesis up another level with physical modelling, which is based on more precise models of strings, reeds, pipes, and other resonators, and they produced a couple of very interesting, very very expensive instruments, before abandoning the tech. Sampling is better for imitating real sounds, and the synthetic sounds you get from physical modelling mostly don't sound all that much more interesting than FM.

We're about due another revolution, because there's so much processor power now a lot of synthesis techniques that used to be impossible are now practical. But culturally there just isn't the interest in experimentation. Electronic sounds live in a few small niches - even with modular - and there's not much of a push to explore outside of them.