One consistent thread of the scholarly communication debate is that many titles--particularly Scientific, Technical, and Medicine titles--are far too expensive, much higher than the rate of inflation. But the assumption from there is that the library needs to trim away at the collection as much as possible, preserving whatever a particular campus deems to be essential. It's always understood that the library will pay the full cost of whatever is ordered or licensed.
Why? As both John Wilbanks and Michael Clarke have pointed out, for reasons of tradition and institutional inertia our scholarly communiucation system has been impervious to the changes that have fundamentally changed the consumer market (Netflix, iTunes, etc.) On the consumer level we've gotten away from the warehouse model (farewell Blockbuster) and in some cases have arrived at the unit of entertainment most people actually care about--the song, not the album.
But libraries still purchase/license complete runs of journals even though a handful of the articles in those journals will have a disproportionate influence. We're buying the album when we could buy the song.
And the faculty in our institutions continue to publish in journals even though it would be just as easy, and far more immediate, to release their work as a blog post. So it seems we're all caught within a self-perpetuating inertial loop that reinforces the norms and values of the print model, just as the Web has made entirely new modes of discourse possible.
For as long as the current system endures, we could at least share the cost with our faculty. Libraries could still make titles available via the A-Z list and lock out unauthorized users of our holdings via EZProxy (one of the most distasteful of our current duties). In essence, we would still manage the ejournal collection without paying the full cost. Why not? If our users want to maintain the status quo they should help pay for it.