We had a serious issue with collisions in the hash values of the lock resource when parallel loading 25 million rows into a 1 billion row table. We were getting unexplained deadlocks that based on the data being loaded shouldn't have occurred.
It was confirmed that this was due to the values held in the table for the composite clustered index key and the algorithm used to generate the lock resource hash value meant duplicates were occuring
The undocumented feature %%lockres%% helped us prove that this was the issue. Without it I doubt we could have proven this issue.