Agent communication should be designed as Byzantine

Agent to agent communication is the fad and we do have a strong opinion on how it should be designed.

Is this feasible, all agents verifying everything all the time?