This reorders lines randomly, so the sequence carries no relationship to the original.
Why deliberately randomise
To sample without bias. Taking the first fifty rows of a sorted list gives fifty rows that share whatever property the sort ordered by. Shuffling first and then taking fifty gives a sample that represents the whole.
Where sampling is needed
Checking a large crawl by hand, reviewing a subset of pages for quality, or picking test cases from a large set.
Randomised order for fair testing
Where the order of presentation could influence a result — reviewing variants, for instance — shuffling removes position as a factor.
Try it: Shuffle Lines Randomly on SeoWolf's Notepad.