> which is guaranteeing that single node failures can be tolerated with zero data loss or loss of availability

We solve this with distributed state machine replication. You don't need multiple writers to solve the single node failure problem. You only need multiple writers for a write throughput scaling problem.

They're separate problems.