I don't see how there is any room for opinion. If you read the article, it opens with an example where it is impossible to determine the correct merge from the given information without diff3.
Bold of you to assume people read articles before commenting.
diff3 has another advantage in the general case as well since it allows you to better understand if something outside of the hunk you're applying also needs to be changed when backporting something.
