I'm trying to determine whether I should use the rowversion column to find data changes for an ETL process that will push updates to a cube. According to Erland Sommarskog, in the post linked to below, "if there are plenty of concurrent updates, it can become very difficult to avoid that you miss changes" using rowversion. Unfortunately I can't find more information about that and while I trust Erland, it would be easier to use rowversion if we could.
http://social.msdn.microsoft.com/Forums/en/transactsql/thread/7f1cea96-7cf4-4feb-97f4-c15095adf449