If you can characterize the latency involved in messaging between domains, it stands to reason that you can provide a reasonable bound on how close both systems

If you can characterize the latency involved in messaging between domains, it stands to reason that you can provide a reasonable bound on how close both systems are to an ideal shared clock source. If it's good enough, who cares if it's perfect?