> I always seem to find an excuse as to why it's not worth continuing, right before actually accomplishing anything with the tool/skill in question. ... As such, I spend many days feeling like the donkey in front of the stack of hay and the pail of water.

I can really relate to that. It's necessary to choose something that is not too far out of the comfort zone. Like choosing something novel is good but the majority of the libraries/tools should not be new. That's key to be able to both start and continue a long time. I'd make sure the first proof-of-concept can be done within a month, afterwards I'd either stop the project or continue. Otherwise it's just too frustrating.

Afterwards you can still build stuff on top. If the problem is open ended but has a reachable first plateau that provides a basic use case this can be a nice "rabbit hole project" that you can spend an infinite amount of time on if that's what you're looking for.