Each title is valuable if it is eventually responsible for gaining its publisher more subscribers; even difficult books that no one ever reads are useful, if they add luster to the list. In general, publishers will be able to support more authors and a wider variety of specialty topics.
Publishers can encourage more people to subscribe by reducing subscription rates as people subscribe. Since computers will be doing the distribution and billing, there is no reason for a fixed rate per subscriber. Since overhead is now largely fixed and marginal costs are near zero, publishers could let subscription rates drop as enrollments rise.
Unfortunately if enrollments drop, rates rise, which could lead to more dropping enrollments, and so on. Perhaps subscription rates should start higher than that necessary for comfortable profit, then drop to the precalculated floor level as enrollments rise. Even if that idea is ruled out, publishers should let heavy readers pay less per book on the principle that subscribers who buy many books will remain subscribers longer than those who do not buy heavily. Today publishers have no way to give preference to the very people they should be targeting most.
The point is for publishers to keep subscription costs low and to increase readership as much as possible, but only to a level that can be maintained indefinitely. After startup, their marginal cost to add a new subscriber will be effectively zero, yet each new subscriber pays the same.