This is a good question @MaxMaxis.
From my of view point of view, right now, trying experiences and invests on alternate hardware (except x86_64 architecture) is too hazardous.
With component shortage, sbc manufacturers are in a in a kind of “one shot” situation with new products: when you can get your hands on a stock of a given chip (eg Amlogic S905Y2 for Radxa zero or Realtek RTD1395 for Banana PI BPI-M), then you’ll go for it. Isn’t your buisness building and selling SBC ?
In this conditions, no one can expect a good software support of all the zynthian hardware ecosystem that is needed: SPI display, I2S compatible DAC, WiringPi library port, up to date basic libraries (some manufacturers distros are still under Debian Stretch !!).
And the 2/3 years old sbc’s like the ones based on the rk3399 chip that are still in stock are overpriced now (but may have a good software support).
The RBPI foundation itself suffers from this component shortage too, and, as you said, Pi4 are hard too find today. But as we’ve have seen earlier with this hardware (the RBPI4, CM4), the Foundation can introduce on the market some totally unexpected boards based on new chips with a reliable software support.
Audio IC’s and even mcp23017 are also heavily impacted, so I think that right now it’s much better to be focused on software development (unfortunately, I can’t really help here 
).
There’s already some devs in that direction (see the API proposal post) and this is pretty cool. Regarding your post but also all the posts about porting Zynthian on a different arch or different distro I think this is the good option: having a good software basis (synth and effect control, their routing, mixer control, snapshots …), for a headless system that should be
- portable on all kind of basic linux distro,
- controlable with MIDI and/or OSC and/or specific zynthian hardware (4 encoders, 4 push buttons)
- extended with any frontend toolkit (like Qt, see post about ZynthBox)
- complete here
Oh ! And : HAPPY NEW YEAR