Yes to the loop join - but that's exactly what is required 😉

Fetching the top or bottom one row per partition is the name of the game here.

Given a straight choice between a WHILE loop and a correlated loop join... 😉

It would be hard to show a real performance difference over a small number of iterations (like the 10 here), but my point is simply that a set-based solution (correlated loop join or no) is a better design.