The only "downside" with clustering is that any methods that operate on multiple keys MUST be on the same cluster node.
This is where you take advantage of the key-hashing. So if you know you're setting a bunch of data for a single user at a time, you can make sure the keys for that user's data are hashed to the same slot.