(1) Yes, this is new and only applies to AIR 30 (and future releases). The behavior is that from now on, devices with OpenGL ES 3 will only support ETC2 with hardware support, ETC1 will use software decoding. Devices with OpenGL ES 2 will use hardware-decoding for ETC2, just like before.
Unfortunately, Adobe didn't put this information into the change log of AIR 30. However, they told me about this via email when I tested the new implementation.
(2) I didn't get a specific reason either, other than that this approach was less problematic on their side — my guess is that it has something to do with changes between ETC1 and ETC2 that couldn't be handled in a clean way in their internal API while supporting both formats at the same time.
I would also have preferred a clean support of both versions, of course. However, I still think it's a step forward, because ETC2 has a much better image quality than ETC1, and devices that support OpenGL ES 3 typically have more memory than those that only support ETC1, anyway.
My tests indicate that compressed ATF textures with ETC2 have exactly the same size as those with ETC1 (to the bit!), but yield a better image quality.
The difference in the file size is solely the result of "-r" compression. Perhaps JPEG-XR-LZMA compression can simply compress better when the image quality is lower ...?
I do understand your issue with file size, of course. 😐
It seems to me that the real issue is that "-r" compression is (a) not usable at the moment because it produces artifacts on some devices and (b) that it yields a lower compression with ETC2 than it did with ETC1.
(3) I don't know, but I'll ask! My guess is that this applies to iOS just the same. Note that all iOS devices targeted nowadays (>= iPhone 5S, >= iPad mini 2, >= iPad Air) support OpenGL ES 3.
(4) PVRTC support hasn't changed, so the answer is the same as with (3).
(6) I know that Adobe is working on it, but they didn't give me any estimate how long it will take.
How exactly are you using your textures? If file size is the main issue, it is feasible to simply stick with ETC1 and software decompression?
In any case, I'll try to get more information from Adobe.