The general advice should be: use PostgreSQL in case you are uncertain what to use. Watch some Youtube video's with Michael Stonebreaker (2014 Turing Award winner) and start getting disillusioned by the NoSQL hype.
Then, try to understand the mess Edgar Codd tried to fix in the '60s and '70s.
You can write a DDP backend which is backed by Postgres. In such a case it ought to feel like Mongo is just serving materialized views of the genuine, consistent data. If you treat the Minimongo data that way—just consistent enough to show an image once—and verify all the writes on the server then you ought to be able to get by.
Have a look at ToroDB (https://github.com/torodb/torodb). It's open source, MongoDB-compatible database which uses PostgreSQL to store data. In a relational way (i.e., no jsonb, no blobs). It's still under heavy development, but worth a look (ToroDB dev here)
Should the general advice just be "store in MongoDB everything that doesn't require consistency and use Postgresql for everything else"?