I've a shard draining – this takes a huge amount of time (see SERVER-14704).
I recently created a new database on the cluster, which ended up on the draining shard as the primary shard.
As it's draining, you expect data to go elsewhere...