I'd say it's rooted in statistics:
1. Take this tire. If you know that there's a high chance it will blow up, then ... the more you sell, the more tires blow up in the wild; the more tires that blow up, the more injured & dead. The more injured & dead - the more likely you are to get caught. The end-to-end equation is "the more tires you sell, the worse your chances to get away with it".
2. Dieselgate: it's different, in that nothing "blows up". But still, the more cars you put on the road, the higher the chance one owner/mechanic will notice it. Maybe your "chance of detection" doesn't increase dramatically with the success of the product (though it definitely does increase), but the impact of it being detected increases linearly with the number of cars sold! This is still a very bad equation, since VW had zero deniability - once the cheat was detected in one car, it could be easily reproduced in all.
It just doesn't seem to work out, from a risk-benefit perspective. Maybe only if you're looking very short term?