I continue to believe that the success of either approach will
depend on whether or not we have an effective binary compatibility
solution. So for me the fine print on my opinion is "assuming we
have a good enough binary compatibility solution".
The Jetty project (now) sees little difference between the approaches: The big-bang approach is not all javax APIs, it's just the EE ones; The incremental approach is not no javax APIs, but a significant number that need some minimal update. The big-bang and incremental are probably closer to each other than they are to the all or nothing extremes.
A key success metric will be keeping existing applications running in either on supported EE8 platforms, combined EE8/EE9 platforms or EE8 binary compatibility on EE9, so that application developers do not feel pushed or hurried into upgrades until they are ready to do so, and that when they do, then the name change will not be a significant hurdle. If we can do this for one round of name changes (big bang), then we can do it for more than one (incremental). If we fail to make the transition a good one, then perhaps big bang is better to get the pain over with, but it will still be a poor outcome.