Add a schemaless database that can index and query JSON objects.
Being able to store your documents as pure JSON provides a huge benefit because you never have to worry about specific data types.
Chris Nicola commented
Support for existing document database would be even better. The bottom line is Table and Blob storage are not even close to sufficient for this need.
Or (better) make it possible for run MongoDB and other no sql databases on Azure, for example graph databases like Neo4J. Not every persistence problem fits into SQL or the "bigtable" type of datastore like Azure table storage. You can't exploit local storage obviously, and Azure Drive is only writable by 1 role instance so that doesn't work either.