I've changed my perception on this a bit. It's only natural to want the latest and greatest that's currently available. However, with how these releases have been going, I think Tesla's approach is often something more like: Let's change X for 10.9 and get a subset of people on that. Then, we'll change Y for 10.10 and get another subset of people on that. We can then compare data for X and Y with the people still on 10.8.1 and see how things are panning out, with the goal of getting the best of all three possibly in 10.11. Meaning, 10.9 and 10.10 may be more experimental and not necessarily the "greatest" in hindsight.
Just my two cents, of course...