They could probably squeeze more depending on their workload patterns. RDBMS' typically optimize for fast/convenient writes. If your write load would be fine with a small increase in latency then you can do a lot of de-normalization so that your reads can avoid using tonnes of joins, aggregates, windows, etc at read-time. Update write path so that you update all of the de-normalized views at write time.
Depending on your read load and application structure you can get a lot more scale with caching.
