Actually I've found that using the "latest and greatest" introduces FAR more bugs to code than using something that has been thoroughly tested in the real world.
As for execution times...it's a case of "lies, damn lies, and statistics".
The code execution times reported by the jQuery folks don't take into account the ever-increasing overhead of jQuery itself. The last version of jQuery where there was a performance increase in real-world production code was 1.4. There was a noticeable drop-off in performance after 1.5, and that's why I cannot ever see using any version of jQuery from 1.6 onwards.