A few points. 

1.  Jim O'Donnell asks: "How large was the universe of academic journals held in American universities in 1995?  How many titles?  Now, how many do we have now?  I pick 1995 as the year in which big deal pricing began more or less and the big publishers started offering e-access."    

That is a great question. Is there any reliable source for this data, broken out by subject area? I can think of  one or two possible sources for this data but it is presented annually.

2.  Danny Kingsley mentions that "..we often frame these discussions [about download use, citations and so on] around the academic use of research papers, and it is true they are the ones creating citations. But in a university there is a huge cohort of use by students which is important to consider. In terms of cancellations, particularly on a large scale, it is likely in the first year the student use won’t be greatly affected - the percentage of papers that were published within the last 12 months put on reading lists for students would be small. But as years progress and access remains cut off this might start to impact on the student experience."

This is a nice point about the need to take student use into consideration. This does drop out of view in many conversations.

I'd take it however in a perhaps different (?) direction and suggest that a contracted journal space is consistent with this point, from the standpoint that journals should assume much more of an integrative and review role than they do now. That can be accomplished with significantly fewer of them.  In instruction sessions for undergraduates, I routinely talk about the great value of review articles in the context of their value for developing background knowledge. They play a key educational role.  But the same point extends to graduate students and researchers new to a field.