First of all, big thanks to @s1ddok and @nader for stepping up.
As I have mentioned before, my own time on this is very limited. I will still be around, but 95% of my time, I am working on a big cocos2d project (not a game), which I hopefully can reveal more about in a couple of months. And hopefully recruit some qualified beta testersfrom this excellent forum : )
Otherwise, I fully agree that the Apple tool chain is the best.
I will never myself use cocos2d-x, not because it is the dark side of cocos2d, but because when I stopped using Windows and C++ 5 years ago, my life quality immediately increased by a factor 10. Yeah, you can use cocos2d-x in xcode ... if you have plenty of time waiting for the compiler to finish, and like buggy debugging.
So I am 100% for cleaning up the code, and focusing on Apple 100%.
I do not believe in a V4. People struggled immensely to upgrade from V2 to V3. Much more than I had anticipated, so it has never made sense to me, to do it all over again.
The main problem with V3, is, that no matter how you pick your nose, compatibility with SpriteBuilder will break. So my suggestion would be, to keep 3.5 99% backwards compatible, but forget about SpriteBuilder. As SpriteBuilder is unlikely to be updated, it simply is too much of a hamstring to continue to keep it working.
As an example:
For nearly a year now, I have been using my own implementation of CCFileUtils. It is 50 lines of easy readable, highly flexible code, which works with new devices. It is 99% backwards compatible with 3.4.9, but it of course does not work with SB and CCB, because some genius at some point figured out that SB should work in mysterious ways.
Compare that to the current CCFileUtils. Does anybody even know how it works?
Bottom line is, that you CAN make major changes to 3.4.9, and maintain backwards compatibility, if you just drop SB support.
If there is to be no x-platform, I would not worry about it. The main issue would be to make sure cocos2d-objc interfaced nicely to Swift code, but I think it already does that. 99% anyways.
So I would, as @s1ddok mentioned, focus on cleaning up, better support for new features, new devices (it is dirt easy), better asset handling, etc.
Finally I would remove effects from the API, as it is sub-standard.