The way the levenshtein algo works is relative to string size, which in the previous sort implementation could lead to unexpected results. Now we normalize the length of all inputs so the scoring is consistent across the data set, resulting in less unexpected sorting results.