But in this case, it actually hampered liveness. The system died because just one server failed.
A simple 1-node failure is trivial to test. There's no excuse for a broken system as a result of a single fail-stop failure.
It takes time to build and test a BFT protocol, as should be evident from the thread. Takes time to go from a synchronous network assumption to an asynchronous network. There's no way that happened in a few days. Caveat investor.