There are no easy methods to kill the damned things. The 0% rollback is actually a documented problem and, IIRC, has an open/active CONNECT item against it.
There are two types of 0% rollbacks... those that are benign and those that eat the processing time of nearly a full CPU. You can just ignore the ones that are NOT using CPU time. Tthe ones that are using scads of CPU are obviously a problem and, as of right now, the only ways to remove them is to either bounce the service or bounce the server.
As a side bar, I try not to kill spids. I send the user an email. Then I find out where they sit and pay them a personal visit to try show them what they're doing wrong. I might even help them with their code. Most of the folks are absolutely grateful.
--Jeff Moden
Change is inevitable... Change for the better is not.