I generally like this approach. Typically my workflow will be to prototype a feature, see where the existing API needs to be extended, and then carve out a two-

I generally like this approach. Typically my workflow will be to prototype a feature, see where the existing API needs to be extended, and then carve out a two-PR sequence; I) refactor the code to make the new change easy, II) make the change. One issue with just submitting the refactor PR on its own is it can be hard to tell if the refactor is sensible without seeing the new usage. I like stacked PRs for this case.