Agreed that "contribute to existing projects" is not a real solution. If you want to solve a particular problem, or set of problems, there will be a finite set of solutions close to what you need. It's entirely possible that all of them suck at some level from architecture to implementation. Additionally, the current maintainers might resist the change of direction that you need, and maintaining your own fork of a large alien codebase might truly be more work than maintaining your own smaller and more familiar codebase. Many would argue that this was the case with distributed filesystems (my own specialty) from approximately 1995-2005, and perhaps even today. Saying that everybody should work on somebody else's project is like saying that everybody should work for somebody else's company. Nope. Somebody has to start the new thing before others can join.

The trick, as you say, is learning to let go. Accept that you'll have other ideas, and maybe one of them will gain more traction. In the course of a long career, maybe even one of those old projects will come back to life because of a shifting context or because working on other things inspired that one extra twist that makes it ten times more valuable. Setting something aside is not the same as throwing it away, and every good developer I know has dozens of ideas that they've set aside. You never know which one will be useful again some day.