It's not a distributed database. It's just a single-node key/value store where all updates are sent to and then read from a Kafka topic, with relational querying and schema management handled by various layered Apache projects.
It's a useful tool/model to follow if your data fits on a single node and you want easy replication across all instances.
This isn't a real product, it's an experiment showing off what you can do.
Kafka is a distributed log, you can use it as replication log if you want and there are tons of production-ready connectors to all kinds of databases already. Anyone who needs non-native replication (perhaps across different databases or sources) should just use those.
Unless your data and the data you are joining are strictly shard-local. But that's also not a distributed join.