I guess it's a semantic disagreement but I would still call this a bug. If they considered this use-case when they designed the system and then decided it was not worth designing for then I suppose it is not a bug, but I find it hard to believe that is the case.
I don't (find it hard to believe). A few chargebacks are nothing compared to adding friction to the signup of the majority of users, financially speaking.
My opinion leans towards - they considered the use-case but ruled it as uncommon and so didn't design for it (or pushed it down the priority queue). This is because it seems that most people in the US (at least those with smart phones) are on some form of contract rather than use a pay-as-you go; it also seems to me that those on contract would most likely retain their number even when they move (within the US).
I understand your point, and I wish I could also find it hard to believe. But there is nothing in the article to suggest that the system is not working as intended. Of course they knew that phone numbers change hands, yet they designed their system to let the number act as both identification and authentication, knowing that it would lead to a certain number of fraudulent charges.