Is this correct?  That, should one or two drives fail in a stretched cluster environment, the cluster fails over to the remote site?  I don't think it is but either way the documentation should read more clearly.