When the defragmentation logic performs a moveChunk, it invalidates the routing information associated to the destination shard but it doesn't invalidate the one associated to the source shard.

This is not a problem right now because the invalidation that it's being done there is just advancing the time in store and it is not actually needed to do it twice. However, this is not optimal for SERVER-64108 because this ticket will invalidate parts of the latest cached chunk manager.