I think this is the natural process, but how you implement it doesn't matter. A lot of GitHub repos use "unlabelled issue" === "a discussion thread". The benefit is that instead of having to search two separate systems, you can just search one (if you can have an aggregate search over both then it really doesn't matter), these two implementations are isomorphic

I don't find them equivalent or isomorphic. You are rolling a generic discussion forum and a work planner/tracker into one. In my experience, such two-in-one solutions rarely work well. Discussions and Issues tend to interfere with each others' UX significantly when there is no explicit boundary. And it looks like Mitchell Hashimoto and a lot of others feel the same way.