I think that's correct in principle, but the important difference is that you don't have to know the priority in advance.
E.g., to implement an Israeli queue using a standard priority queue, you'd have to track the priority of each "friend group" in a separate data structure.
Concrete priority values are also difficult to reuse without accidentally inserting a new group in the middle of the queue - so the values are at the risk of overflowing.
is it true to assume that in some cases, some messages in the queue might never get processed, if any other messages in the queue have an infinite amount of friends?
I'd call it an affinity queue, or perhaps an associative priority queue.