Mango is indeed scalable in both dimensions. There exist users running central instances through cloud services. There are capabilities to add data collection nodes and network them to existing Mango instances, and Mango is highly threaded for vertical scaling.
As far as middleware, we're actively trying to determine what sort of middleware or message bus will be most suitable to Mango's architecture, but there is no uniform API as of yet. That doesn't affect the performance of the server, though.
And for databases we have an implementation of a embedded, NoSQL database under active development. There is also an existing implementation of Cassandra database services, although it has been sitting in the garage for a while and needs a thorough tuning before it's fit for deployment. Existing database solutions, like MySQL or Amazon's cloud database, can be sufficient for terabytes of data, depending on your use case.
Also, as Mango is capable of being deployed in configurations where servers aggregate to database servers, there is basically no limit to the quantity of data you can acquire. These questions fundamentally rely on your use case and the structure of how you intend to use the data. While Mango could function in a swarm configuration, it requires a slightly different set of considerations than a hierarchical aggregation of data.