I would suggest only sound cueing the collisions and pops.
The collision/creation of sphere could be bell-like, maybe a little crystalline sounding. (I am partial to windchimes, after all.) It should have a crisp start, then fade to a low level until the pop. (Or a slight increase in volume before the pop.) The pop could be a second cue, started as the bell is turned off.
If you could vary the rate of playback of a single bell-like cue, you could get many different pitches, creating a very colorful mass of sound.
I can think of a couple different approaches to coding this, but not in 4k! and not before 2012. I made a variant of the Java Clip that allows multiple playbacks, concurrently, AND pitch variation. If you can find a way to adapt the code, you are welcome to it.
You’d probably want to eliminate the stereo, use a cheaper audio resolution, and you’d have to handle going directly to a SourceDataLine output rather than to the AudioMixer output.
Anyway, your game is fascinating to watch, and not every game has to be for brainiac strategizers. There is a definite nice, artistic “generative” aspect that I like a lot.