I have been working with MongoDB since about the time it came about, so some seven years now or thereabouts. Mongo can definitely scale, and it has certain features that allow it to scale for certain purposes better than your average monolithic RDBMS, since you can shard the crap out of the Mongo server and distribute load out the wazoo. Mongo can perform very well in many scenarios, and exceptionally well under specific kinds of read scenarios. Due to it's sharded nature, if you need write guarantee across a majority of shards then that can slow writes down while guaranteeing that when reads happen they read the right stuff.
I am not going to sit here and say Mongo is always better, it has it's uses just as much as any RDBMS has it's uses. Sometimes the scopes overlap, however there are definitely areas where they do not. You do need to use the right tool for the job if you need the best performance out of the tool. Sometimes Mongo is the better choice for certain things (i.e. highly distributed independent reads of objects from well indexed collections), sometimes an RDBMS is better for other things (i.e. complex reporting, large transactional processes). Either kind of database can handle massive volume...I know institutions that use Mongo to store hundreds of billions of objects in collections.
Within the least year or so, MongoDB has gained some features that can greatly expand its capabilities in the realm of reporting, which historically I would say was definitely the domain of the RDBMS. The aggregation pipeline and $lookup stage have expanded the ability of Mongo to efficiently retrieve and reshape large quantities of data, even across a large sharded replicaset, often real-time. This has just improved Mongo's ability to scale, as what once may have required multiple independent requests, each incurring overhead, can now be done with an aggregation in a single call. The performance improvements here can be as many as 10-100 fold depending on the exact nature of the aggregation.
It is also important that you don't make the mistake of assuming that Mongo will scale without effort. You must configure it's servers properly, and put them on the right kind of hardware and networks, such that they will perform optimally. Mongo won't necessarily perform all that well in a single-server environment, whereas an RDBMS certainly could. Mongo could potentially run circles around a classic RDBMS in a highly distributed and sharded environment.
There are many large companies that use Mongo in very high volume, high throughput environments:
Just keep in mind, these companies employ highly skilled Mongo devops and administrators just like any company that had a high throughput RDBMS would employ highly skilled DBAs.