I'd love to tackle this problem using "Big Data" techniques.
It's pretty clear from this story and from things like the Morning277 sockpuppet investigation[1] that the manual style of investigation to so time consuming and slow that it limits the ability of Wikipedia to fix problems like this. While tools like CheckUser[2] help, it seems likely that a lot more of the "legwork" associated with the investigation could also be automated.
At the core, much of the work is network analysis, and that is something computers are better at than humans.
[1] http://www.dailydot.com/lifestyle/wikipedia-sockpuppet-inves...