I was reading this article https://wiki.haskell.org/Correctness_of_short_cut_fusion on the Haskell wiki. It presents an example (2.1.2) where destroy/unfoldr fusion behaves oddly. If I use a lazier definition of unfoldr, then this problem goes away:

Does this mean unfoldr is 'too strict'? Or is there a good reason for not writing it this way (performance, perhaps?)