In the Cluster Edition, a task lock request triggers the cluster lock request, but once the cluster lock request is made, the cluster lock request is independent of task request status.
Suppose, for example, the task requesting the lock leaves before the lock is granted at the cluster level. This could happen if the lock times out or the task is killed. In such a case, the output of sp_lock, for the instance owning the cluster lock, continues to show “-blk”—indicating a blocking process, even though there may be no task waiting for the lock—until the instance owning the cluster lock releases downgrades it.
