Yeesh,
I am very familiar with the way kernel upgrading "works".
Unfortunately, even with all the testing, bugs still get pushed through. It's a fact, which you know since you obviously read up a lot on kernels.
We've actually had problems with some scheduling in certain kernel releases, which were in fact, later on documented bugs. These would have caused outages if put on machines that got heavily hit. Luckily, we don't haphazardly go running off and getting the latest-n-greatest the second we see the post on slashdot. It's called testing.
99 times out of 100 you are correct, upgrading a kernel will cause zero problems, and perhaps fix some you didn't know you had. But when you're pumping thousands of dollars per day through machines, I'm not going to take the chance of fucking that up because of something that may or may not be there.
If you like, I can give you dozens of links to kernel bugs in scheduling, file systems, whatever - that were in final production releases.
Also, security is not the only reasons new kernels are released. New and better hardware support, changes to the scheduler, changes the way VM is handled, whatever are more common than a security problem.
For your own systems, sure go and upgrade the day a kernel is released. You'll be fine. But we actually have people to answer to when that new feature screws with someones specific app. Saying we upgraded them "because it's newer" aint gonna cut it.
To each his own though, I'll let our track record speak for itself.
-Phil