Somewhat Linux-like. You could probably improve it purely from a git perspective by letting subtask dependencies be many-to-many (the commit graph is a dependency graph), but what you have is probably best for your whole Jira workflow.

I get your POV, but I’ve always considered that long-lived branches in the canonical repo (the one in the forge) other than the main one should be directly related to deployable artifacts. Anything else should be short-lived.

There can be experiment on the side that warrants your approach, but the amounts of merge going back and forth would make this hard to investigate (especially when blaming) I would prefer to have one single commit with a message that describe every contribution.