Link bait title aside, I'm a little bored with these "We don't like x, so you shouldn't use it" articles.
The main downside of MongoDB is that it's new. This means less knowledge of best practices, incomplete or missing support in third party integrations, and feature-lacking tools. It also takes a different approach to architecting systems than you would take when using a SQL approach.
There's also the fact that what they were trying to do in Mongo was not what you should do in Mongo. Use a relational or graph database for data that is best represented as a relation or graph.
Nothing like using a hammer to paint a wall and then say you should never us a hammer.
The main downside of MongoDB is that it's new. This means less knowledge of best practices, incomplete or missing support in third party integrations, and feature-lacking tools. It also takes a different approach to architecting systems than you would take when using a SQL approach.