It sounds like your motivation is more about learning new things. If that's the case, lots of small, half-baked projects can definitely be educational.
You're also mentioning projects that take significant ramp-up time before they might start to get a life of their own.
So, that being said, I have a couple of book recommendations:
If what you want to do is learn software development in a growing way, I highly recommend Mastering Software Technique: https://software-technique.com/. It's all about using bursts of time to learn new things as fast as you can. For your situation, it sounds close to what you've been practicing, with just a tiny bit more process.
If what you want to do is actually build something on the side for people other than yourself (which is _not_ a requirement for learning useful things), then I recommend https://jessicaabel.com/growing-gills/.
For me, I've embraced the non-essential nature of my side projects. I've had quite a few over the years, some large, some small. Being able to put one on the back-burner and come back it later has been very handy. Most of my sideprojects these days are ones I can get to "done" in a day or two, and I have a couple longer-running ones that I pick at off and on. My motivation for a particular project varies, some are meant to scratch a _very_ specific itch:
- https://github.com/yumaikas/dirx was basically done in a late night and a morning - https://idea.junglecoder.com/view/idea/274 I've done 3 times, first in bash, then in Go (in a rather late night, to scratch an itch), and then again in Nim (to learn Nim, and compare it against Go).
Some projects I have are a lot longer running. Both PISC <https://pisc.junglecoder.com> (now on indefinite hiatus) and Tabula Scripta <https://github.com/yumaikas/tabulaScripta> have run a lot longer than my typical project. One quality of both is that they have work at quite a few different levels of detail. Am I tired of parsing work? I can work on prototyping a game in PISC. Do I want to make the performance dramatically better for my artificial Fibonacci benchmark? Great time to break out the Go profiler. Do I want to make it way easier to concatenate strings in a stack-based language? Maybe today I figure out how to write an evalbot. (These are _all_ things I did in PISC at different times).
Tabula Scripta is still early days, I haven't had the time to use it for that sort of experimentation, but it will have a similarly broad set of applications and problems to solve.
And, like other people have said, it definitely helps to do you side projects in a different IDE/Language than your work. And if learning is a big motivator, keeping your projects smaller, or making them the sort of things that's easier to break into smaller parts can be very helpful.