Seol mar théacs é seo: The possible and the impossible in multi-agent learning