A whole 22 cents per person per year for a subscription. Very expensive.
It is when you consider that you're paying that for every member of faculty and every student. Not just those in the linguistic department. Those other departments need their own subscriptions. Before you know, you're spending tens - even hundreds - of thousands of dollars on subscriptions.
Um, either a subscription by a library covers all students in an institution, as your first sentence asserts, or it only covers the ones in a given department, as your second and third sentence assert. If the first one is true, the second and third are false.
$3k for a top-notch journal just isn't that much when subscription costs are often in the $10-20k range for other journals.
I do not dispute that publishers make scads of money. I take no stance on whether that money is deserved for the value they provide. The idea, however, that open source publishing can somehow erase all costs is pure delusional fantasy. Publishing can be made to be not-for-profit, but that is no where near the same as free. It costs money to run the journal, and that's beyond just hosting fees. You need a staff, and each person costs on order of $100k in total budget (salary, benefits, overhead). A journal that publishes twelve issues per year needs a full-time staff of six people, exclusive of the editorial board. That money has to come from somewhere. I publish a biennial journal (once every 2 years), and I do it largely on my own in addition to my other responsibilities, which what I base the 6-people estimate upon. What do these people do? Manage the web site, manage the submission process, copy edit accepted manuscripts, typeset accepted manuscripts, fix problems with figures, layout the journal, solicit and manage advertising/grants, make sure that everything is backed up, make sure the payment processing is PCI compliant, manage the business, etc. (If you think that authors can handle the copy editing, typesetting, etc., then you're going to end up with a publication that looks like it was put together by my six year old, won't get read, and won't get referenced. Trust me, I tried that with the first issue of the journal I publish.)
So, if we assume that an annual $600k needs to come from somewhere, there are really only two sources: the submitters and the readers. The traditional model is that readers pay a subscription fee, and submitters pay almost nothing. The open-access model moves the burden from the readers to the submitters, who now need to pay thousands of dollars per published article (12 issues of 10 articles each means $5k per published article). Either way, the money has to come from somewhere. Publishing is not free.
Again, it is entirely possible that the large publishing houses are making a profit that is obscene. I make no assertions on that. But thinking that the costs can go to nearly zero and indefinitely maintain the same level of quality is naive.