In fact, this discussion resurrects a memory of an essay by Dykstra some 20 years ago (in an ACM mag, IIRC) in which he posited that all computer programs should be mathematically provable, and, in fact, such proofs should be the ultimate and preferred form of software QA. Either the program is provable and consistent, or it is not. Since I've done more than my share of embedded programming with poorly or incorrectly documented chipsets that could produce entirely different results on 2 identical systems due solely to variances in capacitor tolerances, I'd have to take issue w/ Dykstra's opinion from a pragmatic perspective, but the notion does bear some examination.

In that light, since Mathematics is as much about proving as it is about doing, I'd think it would be a good qualification (in general) for a software design/architecture position (tho probably too much for a code monkey level position).