Old sonos users don't want new features -- they just want their existing features to keep working. That should not require more resources.
Sounds like plenty for a speaker.
Not if it's running a Spotify client.
Any streaming service is basically http+decode and can operate with a minimum buffer size of probably 16k. Even if you cache a whole song, that is only going to consume 4MB.
32MB of ram was enough to run graphical Unix back in the day.
I got a lot of real work done on a 486, and it already could calculate faster than I could read the results.
I doubt RAM is any issue, but I'm also fairly sure that they went cheap on flash memory and there is no setting higher than -9 for compression :-)