It uses cgroups to contain forking daemons somewhat and detect certain kinds of failure, but it doesn't use them to work out when the daemon's finished starting and is ready, which is what the blog post mainly seems to be about.