One reason why the Harvard Faculty of Arts and Sciences was wise to pre-commit themselves to grant a license to the university in articles they will write in the future is that this license empowers the librarians to seed and to manage the institutional repository in a much more robust way.
The license applies going forward so that at the moment a faculty member finishes the first draft of an article, the university has a license. Any subsequent transfer of copyright to a publisher is subject to this license unless the faculty member requests that the university waive the license with respect to that particular article.
So the librarians at Harvard will no longer have to decipher publisher agreements or the RoMEO/Sherpa list or the OAKlist to know whether an artictle can go up in the IR. Unless the policy has been waived, any article written henceforth by an FAS faculty member can and should go in the IR. (The details of how this will work in practice have to be ironed out, but this motion establishes the legal framework for moving forward.)
[Note, this legal framework (except for the waiver) is exactly how it works with the NIH Federal purpose license (45 CFR 74.36(a)) for any NIH-funded authors.]