In your documentation, you suggest that the user place the middleware returned createActionBuffer after the thunk middleware. Your code does not actually necessitate this, and there are use cases where you may want to delay thunks before state is loaded: for example, if you need data from the stored state to make a network request.