High-availability is always an issue that I am concerned with self-hosting MySQL or PostgreSQL. I am always worry about the fail-over failing, in case any step in the setup was not properly configured. Let me know if someone has some magical formula for easy replication.

I still find CRDB's replication interesting, since all data is replicated 3-way on every commit, and data also auto-rebalances as needed. With some exceptions, as long as you are connected to any node, you should have access to all data.