The most exciting thing a database can do is exactly what you expected, at three in the morning.
I have been woken up by databases. Not metaphorically — actually woken, by a phone call, because the primary had failed over to a replica that was three hours behind, and the application was now serving stale data to real users who noticed. The database in question was exciting. It had a columnar store, a row store, a vector engine, and a query planner that could rewrite your SQL into something the optimizer liked better. It also had a bug where, under exactly the wrong combination of load and timezone, it would silently drop transactions.
I have never been woken up by SQLite. I have never been woken up by PostgreSQL running on a single machine with synchronous_commit = on and a boring backup strategy. The boring databases don’t make for good conference talks. They don’t have launch pages with animated diagrams. But they have something better: they have been running the same query correctly since before I started my career, and they will still be running it after I retire.
The most valuable property of a database is not throughput or latency or the sophistication of its query planner. It is that when you run SELECT balance FROM accounts WHERE id = ? at three in the morning, you get back the right number. Everything else is negotiation.