Byzantine Fault Tolerance in Long-lived Systems
This paper is written by Rodrigous and Liskov.
There are two important liveness aspects in BFT. Fault detection and recovery. With remote attestation, i.e. secure hardware, fault detection is easy but recovery is not possible. With proactive recovery detecting faults is not possible but recovery from certain attacks is possible. Proactive recovery is when each node is monitored by a watch-dog periodically and restarted with reinitialised code upon some failure.
Their approach is to have a CM (config manager) that checks the finger prints. Upon receiving wrong signature or wrong messages they take action.