Robust location specifiers is an old problem within what used to be the open hypermedia systems (open as in capable of integrating existing software applications into a hypermedia system) research community. It is certainly possible to create heuristics (such as this system here) that work for some use cases (and you can make it more robust by adding multiple ways of specifying the location, e.g., the text selection, offsets from beginning and start, DOM traversals from root or nearest IDed element), but the author is wise to set it as a non goal to make this universally applicable. For that to work, continued control and tracking of the surrounding document would be necessary, and for better or for worse that is not the way hypermedia systems evolved.
