Rolling release is favorable in that sense. I prefer dealing with a single problem every other month or so as opposed to many things breaking all at once.
Exactly. It's also much easier to rollback one change with a regression with a rolling release system than it is when that single change with a regression came packaged with hundreds of other functional changes.
Kernel regression in Gentoo? No worries, boot the old one.
Kernel regression in Fedora? No worries, boot the old one... unless this was across an upgrade boundary. Then you get to have fun.
