I don't mean to knock on the pair programming model. It works amazingly for some people, there are cults around it disguised as consultancies and I even had fun

I don't mean to knock on the pair programming model. It works amazingly for some people, there are cults around it disguised as consultancies and I even had fun with a few pair programming exercises. But... does your company actually pair program? If not, how do you know that you are not testing pair programming skills rather than testing whether the person can work on your stuff? For example, when I write code, I write it in iterative drafts. I find that my train of thought breaks when someone points out a trivial piece of syntactically wrong code that I put in there as a filler for stuff that I am going to work on once I have finished iteration 1. I think of it more like solving a math problem where I sketch a picture and then gradually turn it into a sigma/lambda type proof. It is incredibly distracting if someone stops you in the middle of the picture and asks you how this has any connection to the final product.