About the likely factors, from the standpoint of a librarian: Librarians will not cancel the conventional journals if they are used. They will (at least if they are rational) cancel them when they show no use. It is also of course likely that they will cancel them because of cost, especially if the cost per use figure is particularly high.
What good librarians should and do look at, is not primarily the opinions of their users, but the actual beavior of their users. Expensive unused publications get canceled. The most any library can expect is enough money to buy what the users do use, and not also everything they think they are using or think they ought to be using. We do not buy for our own personal reading; we buy as agents for the users to acquire what they need in the format they prefer. This applies no matter what route the progress towards a free system takes. All it needs is for people to make use of it and stop using at least some of the conventional publications. We keep measuring use and many of us are eagerly anticipating the change, but we cannot act until the scientists do.
