But as the relative funding of libraries shrinks to below 2% of universities' budgets and service priorities shift away from physical books, many academic libraries are moving their books into remote storage, effectively closing the stacks once again. Access is limited to effectively a paging service for known items. Because of the loss of serendipity from browsing the shelves, this increases the importance of the metadata for searching, finding, and identifying the right books. The primary source of metadata, of course, is the catalog, particularly for the kind of books that are being relocated. Even when incorporated into discovery systems, the primary source is the bibliographic record.
But, as we all know, library catalogs leave a lot to be desired...OK, they suck (see here, here, and here). They have evolved very little from their beginnings of lists of titles on pages (literal pages and metaphorical pages).
One problem with all of the systems that use book jacket images is that, for many of the titles being moved into remote storage, there are no covers available. These are often older, less popular titles - that's why they are being moved. So the results use a "faux" cover, with the title layered on top. This defeats the purpose of "virtual shelves". It is no better than a list of titles.
I just don't think our catalog systems have the capability to effectively replace the efficiency of locating a section of the shelves and browsing. Perhaps with some combination of Amazon's LookInside, Harvard's StackLife, and the library's rich metadata, we can get a little closer.