I agree that core improvements should be higher priority than anything else, however focusing entirely on rendering optimizations is probably not a good idea. First of all, 3.1 already has lots of improvements in the renderer, especially the lighting and shadow improvements (“optimizations for forward lighting” & “shadow casting” in your list).
For 3.2, which is probably what GSoC 2015 is going to be about, I would like to see see improvements in the art pipeline. This is where a lot of commercial engines have the advantage over jME and it would be great if we could bridge that gap.
For example this kind of work by @david_bernard_31 would be an excellent GSoC project.
Another thing we are sorely lacking in is animation and particle effects (or just special effects in general). I was prototyping a scripting language specifically for special effects and this is something that would be very cool if implemented.