I don't trust OS-level sandboxes to contain malicious processes. System-level sandboxes (KVM, Xen) just might be able to do it but even that is doubtful.
That's why I'd just separate critical functionality into a completely separate CPU, maybe even with a separate network.