This question (https://cogsci.stackexchange.com/q/4777/301) probably is too broad as a normal question, and is not likely to have a single perfect answer. A better model would be to have many contributions that cover the diverse answers. It seems other SE's tag those questions 'big-list' (e.g., https://mathoverflow.net/questions/37151/what-are-the-big-problems-in-probability-theory or https://cstheory.stackexchange.com/q/18968/9192).
I think it is beneficial to have such questions. What is the general policy here?