I certainly cannot speak for the Morgan (or any other library), but several issues can go into our digitizing decisions. Fragility of the item (or, more usually, its binding in the case of materials in our Special Collections) is certainly one of them. While we do not own a top-down scanner like a Kirtas, we can sometimes outsource this kind of scanning. Of course, costs are then incurred, and we would ask the requestor to pay those costs up front it they wanted the scan (and occasionally they do--it's not cheap, though). This is the same model that libraries and archives have used for decades when creating microfilms of items.
Another consideration, especially for us as we have been absorbing most of the costs of digitization ourselves, is simply how labor intensive will a project be. If it's several hundred pages--like an opera score can be--even if it's safe to flatbed we will generally turn down those requests. We simply don't have the manpower available to devote to a single item like that.
A couple of other points: manuscripts are subject to longer copyright duration than published material. I don't know that this is the case with this Sullivan MS, but it's true for many pre-1923 works. Also in the case of manuscript works, depending on how an archive acquired the item, there may be agreements between a donor and institution that preclude the archive making the works available, or only under certain circumstances.
Lastly, it's worth noting that, generally speaking, paper that is very, very old is usually better than more recent paper. Paper from the last quarter of the 19th century through the first half of the 20th is particularly prone to acidic reaction. We've had a program here for a couple of years now looking at scores and books published in the last couple of decades, determining the acidity of the paper, and where necessary sending them out for commercial deacidification before they turn brown.