No rule is free from exceptions I suppose. The distinction I'm trying to draw is between the decision to use QuickSort and the actual implementation - so if someone is thinking ahead and decides that QuickSort is the best option on a given architecture (considering a couple of options) is doing engineering.

Someone who codes up QuickSort is probably, at the time of coding, not making engineering decisions because the decisions inside a QuickSort implementation have largely already been made. This is simply a view on where the dividing line sits between an engineering decision and something else.