This paper received 5 reviews instead of only 3-4 because this paper kind of polarized the reviewers, so I wanted to add extra opinions to reduce the noise in the final decision. The biggest concern was that it focused only on the 2 agents / 2 item case... and we would hope to have a more complete study at the end - even if, we all agree, the problem is way more challenging. Anyway, this is a paper "in the right direction" so we happily suggest acceptance. To be clear, had it be n agents / m cases, the final outcome would certainly even be better for this paper (maybe next year ? but by 2021, the contribution might be more incremental than in 2020).