Quote Originally Posted by strcat View Post
Rawhide will often be tracking the development releases of the version that's eventually going to land in the stable release, so you're exposed to more upstream problems. Most projects still lack good test suites or don't do any testing at all, and the churn of the development branches/releases is painful. The Linux kernel is a pretty good example... hardware-specific regressions are common because it would be very hard to test everything, and it's nice to let other people find these for you in the earlier release candidates and first stable release .

Arch's testing repositories only get the latest stable release of GNOME and the kernel, with the first or second point release being the one moved to the stable repositories. Unstable releases are released a separate gnome-unstable repository. It's not an enormous difference, but I do run into far fewer issues in Arch compared to Debian Sid (probably due to heavy patching) and Rawhide.
There usually aren't many interdependencies between the kernel and the running system. It's usually reasonably OK to run a Rawhide kernel on the current stable release if you want to try a new kernel, or the current stable release kernel on Rawhide if you have a problem with the dev kernel. There are exceptions, and you probably can't go nuts and run 3.2 on Rawhide or something, but broadly speaking it's viable.