As others have said, the costs of distribution are non-trivial, and are almost certainly a loss for JSTOR on these articles. For example the exemplary arxiv.org, which provides a relatively 'simple' hosting service and where the author prepares the full manuscript- they still have annual running costs in the region of $500,000. Even arxiv.org do not seem to have found it easy to secure funding from other universities and institutions which surely find it essential (most costs are borne by Cornell library).
While there is great value in the JSTOR articles (all published before 1923)- in reality very few active researchers will be reading them. Some of the research will be just plain wrong, some corrected or superseded, and if there is any useful stuff left, then it will be readily found in much more modern and useful presentations.
I think the more important battle is to ensure all current research is deposited in open archives (preferably something like arxiv.org). The historical stuff will surely follow...