This was a rather interesting article. But it made me think of two things I think are worth noting.

First, when dealing with complicated queries over large datasets that do not have indexes, it is often faster to create the indexes you need then it is to run the query without the indexes, not to mention that the indexes will then be available for use in the future. This is not universal of course, but I have very frequently found it to be the case.

Next, it struck me as strange that they banned the use of cursors and recursion for an ad hoc query. Of course I recommend avoiding them in all cases where it is practical, but when you say that something is ad hoc and will be run precisely once and you are on a tight timeline so developer time is likely much more valuable than processor time that is one case where it might be worth using one. Even there, I would look for a non-cursor non-recursion solution first, but if I could generate a solution with a cursor faster than I could a truly set based solution then it may be worth considering in a situation like this.