I'm using Mongodb to store Real Estate MLS data. MLS vendors all have their own schema for storing property data that can change on a regular basis. Rather than attempt to map every field we store each property as a document. We index the important fields (price, beds, baths, etc) but all other "metadata" is stored as a Dirty Attribute (in an embedded document).

We are still beta testing and have 100K properties and ~1 million images stored. Query time is faster than our current LAMP site. We will shard based on MLS Vendor as we add more.