QSF first, if needed, then convert and adjust audio in single output.
If you're doing a force recode, and not depending on the intelligent recode logic you can actually do all 3 in a single step. (QSF doesn't have access to some info needed for intelligent recode to work properly)