http://docs.mongodb.org/manual/core/sharded-cluster-shards/

When you deploy a new sharded cluster, the “first” shard becomes the primary shard for all existing databases before enabling sharding. Databases created subsequently may reside on any shard in the cluster.

Let's say you have a server A with database foo (and lots of data), and servers B and C which do not have the database foo. You then set up a sharded cluster and addShards in this order: B, C, A. Once you enable sharding for foo, the primary is server A, even though it's not the first shard.

I tested this on 2.4.8, happy to do more testing if needed.