(The intermediate nodes all see the pre-image but cannot know the exact destination node (unless the next-to-last node knows the last node is a dead end with only one channel)).