With hstore and jsonb fields I don't really see any reason to support mdb. Is it scaling / replication that makes it worthwhile?
Even the docs in mongo tell you what kind of SQL query certain operators perform… I would rather just do SQL.
More often than not I hear the rationalization the no-sql is easier to develop because you don't need schemas. In my opinion that is a horrible anit-pattern.
What I am curious about is the use-cases where mongo actually is the correct choice of pg.
You always need to know what kind of data you are dealing with. If you use mongodb, you no longer have the database as the source of truth for the shape of your data, so you need to maintain this somewhere else, whether it be in mongoose models, in your API layer or just in your frontend. The point is that you aren't solving a problem as much as pushing it around, and usually pushing it around results in it being handled in the wrong place(like the frontend).