From what I read so far, the main reason Linux introduced the 6-year support for LTS versions was Android, because manufacturers modify the kernel for their devices and can't be bothered to maintain these modifications across versions. Well then, I'd really say, screw that. If a) you absolutely need to modify the kernel and b) you can't (be bothered to?) get these changes in a shape that could be upstreamed, then the pain to maintain that is rightfully on you. Of course, in reality in "Android world", many manufacturers will instead put the pain on the customer, having their device run an unsupported kernel, but that's yet a different story.
So, Android aside, I think it's the only sane decision here, maintaining several exact versions in parallel, each for 6 years, creates a lot of work without any benefit for the project.
I still think FreeBSD found the best way to deal with that: Support each "ABI version" for at least 5 years, so minor updates are very unlikely to ever break anything, therefore it's fair game to expect admins to apply them to productive systems regularly. That's a really nice compromise avoiding the huge maintenance burden of an LTS model, but still providing most of the benefit for admins.
And here's what I'm missing so far: What's the impact for Linux? There's always the claim to "never break userspace", still it (rarely) happens. And the in-kernel APIs change all the time anyways, IIRC Linux doesn't even want to support a stable API/ABI there. So my guess would be: There's no thought at all into that direction. After two years, you'll have to upgrade your kernel and "anything could happen".
