It's a very interesting idea. I think it's also worth trying to anticipate how it might fail or be exploited. What better way than intentionally trying to exploit it?

(1) If you're a professor with a big lab, you can assign younger grad students to review and tell them to gift review tokens to older grad students who need to get their paper(s) out to graduate and/or who are really productive with paper-writing and don't want to be bothered with reviewing. Although this is not terribly far from what happens already, at least when it happens now the junior grad student does legitimately gain expertise and loses nothing but time. With tokens, they might also lose a chance at publishing (or have to do more work to earn that right).

(2) It is rare enough already to get a deep, thoughtful review. But if the reviewer is only doing the review to get tokens? How are they going to have any incentive at all to take appropriate amounts of time? The rewarded skill will be figuring out what the editors want to hear and giving it to them with minimum effort, so you can publish your next thing.

(3) Not actually an exploit but a question: what happens with surplus tokens obtained by people who review a lot?

I still think it's probably worth a try. If you manage to get it implemented and tested, please do post an update! Even if you publish the results (Frontiers in Organizational Psychology or something?), a note to Medium readers would be nice.