You need minimum three nodes for high availability, because you need a majority to avoid network partitioning - this sucks. If you’re small scale you might not have three nodes that are good enough to run your control plane. What if instead we could have super-lightweight oracle nodes. Literally all they do is have an API that can be called by the other nodes and replies with which nodes the oracle can see. That way if a node sees that another node is down, it can confirm with the oracle to avoid partitioning!