We are also one of those companies that bought into the promise (or hype) in 2011 and gave MongoDB a spin. After some very successful tests, I had long conversations with my CTO on why we should deploy MongoDB in prod and make it our primary OLTP store. Yes, MongoDB had it's challenges in 1.6 and it required some "faith". But it's 2017 now and MongoDB totally delivered on it’s vision, at least for us.
Since 2012, that's when we upgraded to 2.0, we had 10 min of total downtime. That's 99.999% uptime in 5 years. Year after year, 99.999% uptime. And that 10 min is the total for scheduled (btw, we don't have scheduled downtime for our MongoDB clusters since there is no need for that) and un-scheduled downtime. We never take the service down. We perform all maintenance operations during regular business hours with no impact to end-users at all. So much about reliability!
MongoDB as a product has bugs. What a surprise, what software product doesn't. But it's about you community (we are running the community edition) supports the product. We recently discovered an issue as well causing the primary to crash. As designed, another member in the repl set took over and there was no downtime at all. Even though we don't have a support contract, the issue was acknowledged by MongoDB in hours and fixed in a couple of days. I wish we had that type of support for products where we do have a support contract!
And last but not least, if you can't figure out how to successfully use MongoDB, for god sake, take some free training classes on MongoDB University.
MongoDB might not be a good fit for every use-case, but it's already an awesome platform for an incredible number of applications. Polyglot environments is the keyword here. SQL has it’s advantages and disadvantages and so does NoSQL. Use the best tool for the problem. One size fits all doesn't work anymore and MongoDB is clearly the best and most popular document store on the market. And even better, other platforms like postgres have acknowledged that the idea of storing JSON has its merits and have implemented its own version of that idea. However, IMO, if your use-case is a good fit for a document store, use MongoDB. We have done so and couldn't be happier with that decision.