You are assuming that these requirements cannot be met by a "thawed" kernel; I disagree; and I do think I have some perspective. (And that is ignoring the small detail that the "frozen" enterprise kernels have substantial change rates as well.)
The main reason why enterprise distributions exist is that change causes fear in humans, whether justified or not. There are several ways of resolving that; one is to avoid/hide the change (which is what enterprise distributions do - "ohhh, version number is static, must be stable").
The other is to make the risk/cost associated with change drop below the benefits of the newer versions. There is no reason why "upstream tip" should be of lower quality or stability than an enterprise distribution. (That it sometimes, indeed, _is_ of lower quality perhaps tells us something important. But that can be addressed.)
Mind, the first model makes good money; and it satisfies a real and justified need the customers have - managing risk. But it is not the only possible solution. While you won't be able to sell the second to everyone, nor that it would work immediately, or that I have all the answers.
But it's one of these things that I'd love to investigate more. One of these days I'm going to shake a tree until a business angel falls out ... ;-)