> Process per connection is pretty easy to accidentally run into, even at small scale. So now you need to manage another piece of infrastructure to deal with it

> Process per connection is pretty easy to accidentally run into, even at small scale. So now you need to manage another piece of infrastructure to deal with it. Most places I saw this as an issue, are where developers think that by tweaking the number of connections will give them a linear boost in performance. Those are the same people that think adding more writers in RWLock will improve writing performance. I agree that it's easy to run into and pretty silly concurrency pattern for today's time. At the same time, it's just a thing you need to be aware of when using PostgreSQL and design your service with that in mind.