Thanks for this fascinating article, Molly. I assigned an excerpt from it to my 1L property students long before I knew this mobblog would take place, and due to a fortuitous alignment of the syllabus, we had a chance to begin discussing it on Friday (along with an excerpt from Glen Robinson's Chicago Law Review piece). So my students will be reading along and, I hope, chiming in.
I think you are exactly right to draw distinctions among different flavors of "new servitudes" and to distinguish among problematic features of those servitudes (in your taxonomy, "notice and information costs," "the problem of the future," and "harmful externalities"). I just have a couple of suggestions on how you might go even further on those fronts.
First, I wonder if you could gain even more traction from land servitude analogies by considering which of these servitude forms most closely resembles each of the "new servitudes" you discuss. For example, the Microsoft EULA sounds like a negative easement in gross that benefits Microsoft and burdens the end user and her successors. It works much like a conservation easement, in that everyone who comes into contact with the property is foreclosed from doing certain things with or to it, and may present some of the same concerns. The Creative Commons and GPL examples, in contrast, look much more like the covenants in common interest communities (CICs) that reciprocally burden and benefit each parcel-holder. Each person enmeshed in the overall scheme receives a benefit (access to the products of others) but also is subjected to a burden (giving reciprocal access to others). The large number of benefited parties who must release each burdened party creates the anticommons-like dynamic you describe; similar problems have arisen in CICs. So there may be more to learn by looking at the problems and solutions that are associated with the closest analogues for each "new servitude" type.
Second, although your taxonomy does this to some extent, it may be helpful to distinguish more sharply between two kinds of concerns about servitudes: problems that are uniquely presented by servitudes "running" to other users (who may be remote in time and space), and public policy concerns that exist independent of these concerns about running. With land servitudes, the first category of problems are tackled through running requirements and doctrines that provide, under certain circumstances, for modification or nonenforcement (or damages-only remedies, as Stewart Sterk notes in his comment). Broader-based measures are used to address the second category of problems. For example, the Fair Housing Act places side constraints on servitudes that run with the land, but it would also limit in exactly the same fashion and for exactly the same reasons non-running discriminatory agreements among individual neighbors.
Running itself may be an issue in the case of servitudes attached to Creative Commons and GPL licenses, in that it expands the web of reciprocally-bound parties. And at least one concern that you raise about EULAs implicates running as such: the worry that restrictions affixed to obsolete software will hamper its use for education and research long after the software has ceased to hold significant commercial value. [A quick aside on this point: Might one possibility be a sunset on the enforceability of licensing agreements once the software in question is no longer available for purchase from the manufacturer or its successors? As long as the manufacturer continues to offer the product for sale, the use of servitudes could actually advance educational and research uses, for a reason noted by Glen Robinson: servitudes make it feasible for manufacturers to engage in price discrimination].
Other concerns with EULAs seem conceptually removed from questions of "running." For example, the possibility that EULAs will interfere with fair use rights seems like a substantive objection to the limits themselves, even as applied to the first purchasers of the product. If our real objection is to the limitation itself and not to the fact that it runs, we need a different (and broader-based) response. Otherwise, manufacturers could shift to non-running substitutes with the same substantive effects. Consider what Glen Robinson terms "hardwired servitudes" – electronic products that self-destruct after some number of days or uses, or otherwise directly prevent any user from breaking the original license agreement. These built-in constraints might be viewed as problematic, but not because of the time and remoteness concerns we would typically associate with running servitudes.
An analogy may help here. Suppose a buffet restaurant rations access to its food with tickets that are issued only after a purchaser agrees to the following: (1) the ticket can be used for no more than five separate trips through the buffet line; (2) no containers for carrying food may be used other than the plates supplied by the restaurant; and (3) these restrictions "run with the ticket"; if a ticket is transferred, the trip count does not reset, and the "no outside container" rule remains in force. After observing the buffet's operation, we might well conclude that the customers are getting an unfair deal. Perhaps the supplied plates feature a clever dome-like surface that precludes the efficient piling of food, or the restocking of the buffet is strategically staggered so that people can access only a small fraction of the ostensibly available foods on each trip through the line. But because these are not, fundamentally, problems with the restrictions running to other diners, any response to them should be direct and broad-based (say, though prohibitions on deceptive buffet practices). The running feature is indeed essential to the restaurant's current business model, but that business model (and the running servitudes necessary to sustain it) may turn out to be socially valuable once the substantive problems with the buffet's serving practices are addressed.
Thanks again for the paper, Molly, and for participating in this online event!