If it makes you feel any better, I don't think that paper refutes your claim through any general law. It just enumerates a bunch of side effects of current virt

If it makes you feel any better, I don't think that paper refutes your claim through any general law. It just enumerates a bunch of side effects of current virtualization approaches, all of which are focused on performance/isolation. It's not inconceivable that someone could develop virtualization software that offers a significantly slower clock speed processor to the VM to give it plenty of time to do things in the luxurious gaps without giving away timing information.