Low latency query execution and it scales down well to small datasets, are my reasons for really liking it. Top throughout when scanning TB+ of data on multiple nodes isn't my use case; it's answering a query in web request response times over a few hundred million rows of data.

The low latency to query execution is really nice.