We experienced some significant difficulties with sharding; the automatic methods of doing so only seemed to shard a single-digit percentage of our data. We've also encountered some wildly unexpected issues with master/slave replication and related nomination procedures.

You're right that this post is vague with regard to those details; they would be a good candidate for a future blog post, but the desired takeaway from this one is that we're quite pleased with the performance and scalability that Riak provides.