Interesting that the graph goes up to 10k connections without any cliff in throughput, I wonder what use cases people have beyond that. Or maybe the issue with idling connections is memory usage?

Yeah, the absence of a sudden cliff is very nice. I think we fixed the main causes back in ~9.5. But the gradient at the end is pretty clear, and even at 10k the throughput is already less than 50% of the max. And it's dropping faster and faster.

It's also important to keep in mind this is an extremely simple workload, essentially just read-only point queries, on about 3GB of data.

Postgres is improving rapidly. I started developing my own DB a few years ago, but decided to do something else. One of the main reasons was because of the progress this DB is making on all fronts.