SQL and NoSQL in the same project

I would like to know opinion of knowledgeable people about crossing in a single project SQL solution (MySQL) and NoSQL (MongoDB). On the one hand, somehow not on Feng Shui to use both at once. But on the other hand, separately each decision is not fully satisfy the needs.

On the one hand the project it is necessary to use a very large number of documents with variable structure (different number of fields (which is mandatory)) and different size (from 200 bytes up to 200 kilobytes on average) and this requires a lot of speed work with them. On this MongoDB is very well suited for this. It is also quite easy cluster to increase speed and failover (Replica Sets + Sharding)

On the other hand in the project you want to use a lot of clearly structured data that are very closely related to each other. And store them in a nonrelational database is simply not feasible and quite uncomfortable to extract information(restore relations may require hundreds of subqueries). And all that is connected with removal, will generally cause a lot of inconvenience. At the same time, these data — minimum quantity (in comparison with the number of "documents"). So the whole work can be organized through a single MySQL server.

If you only use Mongo, it will cause a lot of problems when working with related data.
If you only use MySQL, you will have problems with scalability, speed and data storage

Who believes if you want to use both approaches at once, or to come up with anything else? And has anyone used these in practice?

PS the Project is a normal web service but it must withstand a heavy load and should have a good horizontal scalability.