I commented here earlier, not without frustration, about a pair of researchers who built and abandoned a disciplinary repository. I was particularly annoyed that they seemed to have done this purely for self-aggrandizement, apparently feeling no particular attachment to the resulting repository.
Such as they should not open repositories. Neither they nor any service they offer is trustworthy. I hope that's uncontroversial. Unfortunately, even vastly better intentions than that don't guarantee the sustainability of the result, even in the short term.
The Mana'o anthropology repository, started by Dr. Alex Golub on the traditional server-in-the-basement that is the origin of many a worthwhile project, has been encountering significant technical difficulties. Dr. Golub is no longer able to maintain it, and is looking for some way to hand it off.
In my reasonably well-informed opinion, any such one-person effort will need a rescue at some point. If nothing else, people die unexpectedly! Dr. Golub's mistake isn't that he wound up needing a rescue; it's that he didn't anticipate and plan for a handoff from the beginning.
It isn't just Mana'o, I'm afraid. How many disciplinary and institutional repositories have done succession planning? If not, why not? Do it. Now. It is flagrantly irresponsible not to.
Scholarly societies have the best fit with the disciplines, of course, but many who might otherwise accomplish rescues are hamstrung by the need for anything requiring effort to pay for itself or even make money. AAA won't be picking up Mana'o, I confidently predict based on their track record vis-a-vis open access.
Librarianship's continuing error, as I pointed out in the post I linked above, is that we have no infrastructure or plan for accomplishing these rescues, which (I anticipate) will continue to be necessary and may even accelerate in the coming years. Institution-based efforts such as IRs have the technical and human-resource capacity to pick up the slack; what they don't have is policy that allows them to, and coordination to notice work that needs doing and parcel it out appropriately.
As for institutional IT, which might be another natural place to look—they, too, have no policy mandate to address needs originating outside the institution.
How does this relate to data? Well, the problems are the same, really. One-researcher or one-lab IT infrastructures live on a razor's edge; one missed grant may kill them. They hardly ever consider succession planning; worst-case, their IT people (usually wrongly) believe that whatever they're doing is perfectly adequate and will not accept gentle correction.
What this suggests to me, among other things, is that passive data collection is inadequate as a data-repository population model. (Not a surprise, I'm sure; we tried that with IRs and it failed.) Someone needs to go out there and find the good stuff, then open the conversation about how best to keep it.
It also suggests that we need to open a discussion of this issue in cross-institutional fora. CNI, ARL, Educause, JISC, where are you?