What is the scope of definition of probabilistic safety in blockchain/distributed consensus? Could we say BFT protocol preserves only probabilistic safety because the adversary is upperbounded? TLDR: my answer is no, for definition of BFT protocol already states it should be permissioned. Some of the context and reasons below: Longest chain protocol in blockchain guarantee only probabilistic safety. Its safety fail in asynchronous network model; BFT (Byzantine Fault Tolerance) protocol, however, preserves deterministic safety in whatever network model. We say longest chain protocol guarantees probabilistic, for longest chain protocol is unable to foresee which network model are we in. We might be, fortunately, in a synchronized network model, but also an asynchronous one. This depends on luck. BFT is for deterministic safety. That’s because, to safely run BFT protocol, the proportion of adversary must be under certain boundary in comparison with the number of total participating node