Help Olof and welcome back. Glad that you’re interested in using Elk audio OS. Let me try to unpack your questions one at a time.
Boot time is an issue for all linux based devices, but it iis still pretty short compared to a desktop computer. It should def be possible to have sound running in 5-6 seconds if loading the audio code isn’t too heavy. See a longer discussion here Startup & Shutdown
It’s possible to optimise further, but then you need to customise your image and really remove everything you don’t need.
I’m not sure what you mean with class A1 and A2 here.
We currently don’t support RPi5 and we can’t provide a timeline for when, so I would recommend going for a RPi4. It is still pretty damn powerful, so I have no doubts it should be sufficient. RPi 3, 4 or 5 would likely not make difference for latency though.
Some notes on dev environment:
Building on the board works fine if you’re using a dev image. I do that myself sometimes. Though for a large codebase, building on a RPi is pretty slow. If you’re on a linux or Mac dev machine, sshfs works well for mirroring the filesystem on the board on your local filesystem. Using that it is very quick to make edits to the code and recompile on the board.
Any app running natively on an Elk device would need to be cross-compiled for that architecture, not just audio apps. Raspa though needs specific libraries and would be hard to get running on a non Elk OS.
If you want to build your own sound engine and not use sushi, I would suggest you to build that app with support both for raspa and another audio framework (like jack or portaudio) so you can test and run on a non Elk-device too. This is what we did with sushi and enabled us to do a lot of development without an Elk board.