yeah ... in the early days of mongo it was pitched as a replacement for relational because a) it had faster writes without those pesky joins and constraints, and b) because it's too hard to model data, so chances are it's just unstructured. the onslaught of "relational is dead" and "joins are annoying" was tremendous

Plus there's a good chance they didn't know exactly what they needed to build when starting out.

Nonsense. Choosing the correct database isn't "premature optimization", it's the bare minimum of being a competent programmer. Why do we never have time to do it right, but we always have time to do it over?

It's generally faster to get started with Mongo.

If you're talking about a startup, you might not even get to the let's do it over stage. You might have six months to just hack something out so you can raise more funding, and in that case Mongo is a great choice.

The needs of your business might also just evolve over time.

Postgres and MySQL are easy to hack something together with, too.

That's assuming you're experienced with them, if you know next to nothing about databases, Mongo is a much better choice. In this case it looks like when they got started the team just didn't have all that much experience with databases.