Good point and a lot of wisdom in what you say. I agree empiricism and historical facts on previous delivery are the best estimates. If a team normally can put three stories to production per sprint, you have an average that is probably correct.

I agree as well with your point to SD and no manager wanting to have that work ;)

What I implicitly also try to convey is that if you create estimates and promises on what to deliver, you focus on the output and not the outcome. The focus should be on if you provide value and welcome change late in the process. By shifting from focusing on estimates to goals, you become more flexible and focus on what matters rather than how much you produce...