Interesting. I'm currently working MongoDB into a couple of projects I have going at work (they're not really outward facing except for potentially one generated view).
For these projects it seemed to make a ton of sense given the rest of the stack, the nature of the projects, the deadlines I'm facing, and I've really loved working with it so far.
The simple fact is mongoDB fails to have sufficient durability guarantees for business data when compared to a traditional RDBMS. That is one of the reasons why its so fast in comparison. A power loss event during data writes can and will lead to data loss.
I thankfully don't have to worry about power loss (large company, teams dedicated to critical systems infrastructure in a big way), and a cron job can handle any archival concerns in the projects I'm running here.
It does sound like they've improved much since 2011.
MongoDB 4.x also brings transactions to the table, which further reduces the vulnerability to inconsistent data. Data loss is far less of a problem than inconsistency due to partially completed operations and data damage.
For these projects it seemed to make a ton of sense given the rest of the stack, the nature of the projects, the deadlines I'm facing, and I've really loved working with it so far.
Why did you end up moving off of it?