I can relate to your situation. While I have not completely figured it out, one thing I realized is that, even though I shelve many of my side projects, occasionally I come back to some of them, with fresh enthusiasm and move it forward another step or two (before loosing enthusiasm again). I realized that given enough "cycles" at least some of them are bound to be completed.

So a new approach I am trying is the "GPU Pipeline of side projects". First, I choose to trim my list of side projects to a few important ones (4 is a number I like, so four). I work on them in a parallel pipelined fashion. So I have one that's more advanced, one that is being setup, one that is being fleshed out, one still in ideation phase (you get the idea).

Secondly, I realize that when building things, the novelty can come from "recursing down to the details". For example, setup a AWS pipeline for code deployment would mean, learning how to setup an Ec2 deployment, then a Chef or Puppet instance, then easy config management of them, best way to document our setup (gitbook etc) and so on.... So if we use our boundless curiosity, which is both a strength and a weakness, to look a little deeper, we will find enough novelty to keep going even in a "single" project.

What I realized is that one single project is like a LISP S-Expression. (project). You (eval project), and it returns (list sub-project1 sub-project2 interesting-research3 wonderful-idea4) All you have to do is keep eval'uating the sub-expressions till you hit an "atom" of truth to share with the world.