One database server can handle maybe queries per second. What if you need ?
Read replicas: Copy data from the primary (leader) to secondaries (followers). Writes go to the primary. Reads can go to replicas.
With read replicas, you handle x the read traffic.
Trade-off: Replication lag. Replicas might be milliseconds to seconds behind the primary. You might read stale data.
Use read replicas for read-heavy workloads where slight staleness is acceptable. A user viewing their own tweet can hit the primary. Others can hit replicas.