Interesting! My intuition was wrong. I neglected to fully appreciate that the list is memoised.
What's happening, if I'm not mistaken, is that the unevaluated tail of the list is at all times a thunk that holds a reference to the cons cell holding the previous list item. Hence this is more like `iterate (+1) 1` than it seems at first glance.
I am fairly certain lists in Haskell are just normal singly linked lists with pointers pointing to the next item in the list. There's no need to maintain a reference to the previous cell. The last cell in any infinite list (or any finite list that hasn't been fully evaluated) will be an unevaluated thunk but that's not really observable to the program/programmer.
What's happening, if I'm not mistaken, is that the unevaluated tail of the list is at all times a thunk that holds a reference to the cons cell holding the previous list item. Hence this is more like `iterate (+1) 1` than it seems at first glance.