Bike porn

What is wrong with a screw on sprocket by the way?

Engineers don't transmit torque through screw threads, it's just wrong. It only works at all on bicycles because the the thread in question is huge. A spline sized to transmit the same torque could easily be less than half the diameter. In fact, it is; look at the input thread and the output spline of a Surly Fixxer. As it turns out, the thread is not in too much danger as long as the transmission ratio is quite large, say >2:1 for lightweight hill climbers or >3:1 for burly track sprinters, and it equally turns out that efficient ½" pitch roller chain drives at bicycle loads and speeds don't demand smaller sprockets than can be accommodated over a 1.370" pilot bore. But it's still wrong.

I always imagined that if the stays were connected to the rear of the seat lug (like on some of the examples posted previously and that engineers here don't like)

Having the stays abutting the seat lug is good, using them as part of the seat post clamping mechanism is bad.

And when you say "famous names making mistakes", would you really put Ritchey and Roberts in that category?

Yes. Engineering pays no respect to fame, or previous good work. Every engineering choice is assessed on its own merits.