Great post! This problem reminds me of Google interview questions. However, I would like to elaborate on the probability discussion of pixel failure. If p is the probability of all pixels working for the 25-inch TV, then 1-p^4 would be the probability of at least one pixel dead on the 50-inch TV. Therefore, the chance of at least one dead pixel actually increase by (1+p+p2+p3) times. However, with modern technology, p is very close to 1 and so even though this probability increases by almost 4 times, it's still negligible.
RE: Scaling - Why Ants Bounce But Elephants Go Splat