To be fair: these are three very different kinds of store, designed to meet the needs of very different workloads, and trying to solve very different problem sets.
Not the way I see them used. HBase is quite versatile, Facebook's HydraBase is a good example of doing things people think they need Cassandra for https://code.facebook.com/posts/321111638043166/hydrabase-th.... It's also telling they as the creator abandoned Cassandra, but C* has some very niche conceivable uses (eager replication cache for metadata?). Mongo is just stupid.
