If I recall correctly I came across this syntax with help from Erland Sommarskog. I was looking for a way to remove duplicates from a table with over 100 million rows. As it turned out, this was the only method I could use that would perform well enough given the amount of cpu and ram, that would even do the job.