This is speculation, but perhaps there's a game-theory issue here. If IEEE loses money by people pirating (and you can apply the usual counter-arguments about whether the pirateer would have gone on to purchase it anyway) rather than paying then publishing costs increase to cover that loss, and therefore hurting the authors who must cover that increase in future if they want to publish in the journal again.

A better situation is where the preprint or open access version is available.