Comment by prescriptivist

Exactly, if you are doing copilot driven dev, then you are in a feedback loop with the agent and it's easier to have multiplexed ongoing "conversations" with the agent across discrete checkouts of the same codebase than it is to ask an agent to make a change, commit, checkout another branch, wait for the context to update/become aware, ask for a change, agree to it, commit, then switch back to another context.

It's the kind of thing git worktrees was made for.