Thanks for setting up the benchmark in the first place, it has been fun seeing the performance of the different approaches.
Nested loops don't scale well. When there were only 2 items for the outer loop it was really fast. I've seen folks here toss around the Big O notation and although I hadn't heard of it before Perlmonks I figured out they were talking about linear growth as opposed to exponential or some other function for the execution time related to the size of the data set.
The nested loop approach seems to be O(n^2) but I'm still learning so hopefully someone will weigh in if I'm off base.