It doesn't have to be that expensive. They could start with a small scale demonstrator that could fly on Falcon 9 just to explore feasibility and make the mist

It doesn't have to be that expensive. They could start with a small scale demonstrator that could fly on Falcon 9 just to explore feasibility and make the mistakes at less cost. Or once Starship is operational they could fly several different small scale prototypes just to test them in one go, and they could even do this on an early flight that they wouldn't use to launch commercial payloads. Still expensive, because it might need two flights for them to get the confidence they need to go with one design. But not as expensive as "the expensive way".