thanks for the feedback. You are right in that automated testing alone is not good enough for a final release for several reasons (which I also mentioned in the talk).
a) is Qemu/KVM only emulates one certain combination of hardware, so 99% of drivers will not be tested by it.
b) see Dijkstra's quote above - there is a huge number of possible code pathes in most programs and one test-run can only exercise one such path.
c) there are also a big number of packages in modern distributions and few come with their own test-suites
d) software (especially open-source) gets updated a lot and keeping tests for it in a working state can be quite some effort

The automated testing is intended to save developers/testers/users from obvious blocker bugs in the unstable-release-of-the-day (named openSUSE-Factory / Fedora-Rawhide or Debian-sid).

Ciao
Bernhard M. Wiedemann