Withdraw permission to publish an article in their journal that has already been distributed as a preprint. Many journals in biology do that now -- they consider it publishing an article twice, which is a big no no ("self plagiarism"). This is beginning to change and hopefully preprints will be accepted in biology as they are in other fields.
It's little appreciated, but journals are in competition with each other for good papers, too. Editors at top journals are out at conferences trying to find the next submissions and handing out business cards. Nature editors are pissed when a hot paper goes to Science instead (savvy PIs exploit this dynamic). In this kind of 'market', a publisher who unilaterally rejects preprinted papers is putting themselves out of the running for those papers.
preprints are free but when they are added to an issue of a journal, that issue might not be free. so journal X agrees to free preprints (rough drafts) and then charges for the actual issue. if the library drops subscription to the journal, maybe the journal/society/publisher pulls the library's permission to use the preprints.
> If university libraries drop their costly journal subscriptions in favor of free preprints, journals may well withdraw permission to use them
withdraw permission to do what exactly, and enforced how?