Not to take credit away from Andrew for his ideas and writing, because at least he came up with the idea and wrote about it, but I don't understand how that idea of Jepsen style testing of package managers is a novel idea. Like... what testing would you want to do if you were building a package manager?