If you excuse the buzzword, data warehouses are built for this very reason. So you are right in the sense that cache invalidation and denormalization are problems in sql based large scale systems, but they are largely solved problems whereas in mongodb you would need to do this work manually.