The point is, you can backup your system and try CSSU without fear by taking a snapshot, then installing and trying it. Yes, installing a PR1.2 snapshot after updating your kernel will cause issues, because it can't find the modules it wants on the device. That has more to do with updating the kernel than BackupMenu, and shows that while KP can be uninstalled, it can cause issues on uninstall (including breaking things) if you don't use it's uninstall app. You can also brick your device by uninstalling KP with HAM or apt-get, requiring a kernel re-flash from a PC to make it right again.
Yes... It's almost identical language to most things that have a stable and testing branch. What's in stable has been tested and is working well, where what's in testing may in fact break things from time to time. It's the same warning on extras-devel and extras-testing vs extras. (You know, extras-devel/testing, where you got KP from...)
What are you talking about? "including a fix that results in the same"? I can't parse what you're trying to say here. What bug are you referring to?
As for "magical bugfixes", there's nothing "magical" about them. There are several, documented along with check-ins and code. Things are far more stable IMHO than under PR1.3. And I'm not alone in that assertion.
Really? To quote others, where is that at? Can you show me a post or a patch that show that? While possible, I somehow doubt that a kernel patch fixed a leak in a user-world program like hildon-desktop. Far more likely it was a library or a direct change to hildon itself (via CSSU or one of the pre-CSSU releases of changes made by MAG). I still don't get why you trust the community to change your kernel, but not some user/system apps. It would seem there's far more risk in accepting kernel patches (which are not labeled as stable). It's like trusting your neighbor to change your engine out, but not paint your car. Sounds a little backwards.