I've done some Googling and it seems to be a bug introduced by SQL Server 2008 R2 and fixed in SQL Server 2012. But that's the extent of what I'm finding. Any chance someone here has experience with this and how to prevent it from occuring in SQL Server 2008 R2?
One suggestion was to reduce the polling interval on SQL Agent in the cluster resource Advanced Properties. Other than that there doesn't seem to be a fix yet.