I dislike MongoDB, but if I had to try to summarize objectively:
1) It is marketed very heavily, including to use cases to which it is poorly suited
2) It can be very complicated to run in production, I think because it actually wasn't originally designed to be distributed.
3) It has a history of data loss. I think this is because it generally favors performance over reliability, combined with #1 and #2
I have run it in production, thrown over the fence by some dumb devs who believed the lies that "you don't need a DBA". Well you need a full time team and a lot of hardware, and that's comparing it to Oracle.
What about other NoSQL solutions?
MOST other NoSQL solutions actually do things like treat their index differently than their bulk data so that it isn't accidentally swapped out, ruining performance. Many also do important things like supplying durable writes. (Your data may not be consistent at any one given moment, but it'll get there.)