[mmt]

> I’ve come to think that people should just start with Postgres and only pick something else if Postgres doesn’t work out

Even that standard results in people giving up on PG too early. It's possible to get it to continue working out, but it can require effort beyond a naive implementation and often more hardware knowledge than is common any more.

The other options can make naive implementations much easier, at least initially, and on almost any hardware. Of course, that usually ends up costing far more, even at modest scale.