I talked with a Xen developer and came to these conclusions:
- Speculative taint tracking provides complete protection against speculative attacks. This is sufficient to prevent leakage of cryptographic key material, even in fully dynamic systems. Furthermore, it is compatible with fast context switches between protection domains.
- Full time partitioning eliminates all timing channels, but it is possible only in fully static systems, which severely limits its applicability.
- Time protection without time partitioning does _not_ fully prevent Spectre v1 attacks, and still imposes a large penalty on protection domain switches.
Additionally, I am almost certain that:
- On properly designed hardware, both time protection and speculative taint tracking can be enabled and disabled by systems software.
- Time protection and speculative taint tracking are not mutually exclusive. A cloud provider might use time partitioning to partition different customers from each other, while guest OSs use speculative taint tracking to protect different processes from each other.
In short, time protection is excellent, but it is not a sufficient mechanism for general-purpose computing. Speculative taint tracking is a different mechanism that is applicable to many more workloads, and which provides complete protection against speculative attacks. Both mechanisms can be used together depending on system security policy.