I've made large joins faster in MySQL by adding gratuitous order by to derived table queries, which lead to probing the index in the join in index order instead

I've made large joins faster in MySQL by adding gratuitous order by to derived table queries, which lead to probing the index in the join in index order instead of a different, more random, order. The cost of the extra redundant sort was less than the cost of the random btree traversals.