> The first time you build a physical rocket and test sending it up, it's almost certain to fail.
That's not true. Many launch vehicles have succeeded on their first flight:
Titan II: Apr 8, 1964 (Gemini 1)
Saturn I (Block 1): Oct 27, 1961 (SA-1; only 1st stage active, 2nd and 3rd stage were present but disabled)
Saturn I (Block 2): Jan 29, 1964 (SA-2; only 1st and 2nd stage active; 3rd stage was present but disabled)
Saturn IB: Feb 26, 1966 (AS-201; first Saturn 1 flight with all 3 stages active)
Saturn V: Nov 9, 1967 (Apollo 4)
Space Shuttle: Apr 12, 1981 (STS-1, Columbia)
Atlas V: Aug 21, 2002
Falcon 9: Jun 4, 2010 (Dragon Spacecraft Qualification Unit test; the first stage disintegrated on re-entry, but that's irrelevant to success of the payload)
Falcon Heavy: Feb 6, 2018 (test flight; center core recovery failed, although again that arguably doesn't count)
Vulcan Centaur: Jan 8, 2024 (Peregrine lunar lander; the lander failed, but for reasons unrelated to the launch vehicle)
Indeed, I'd even say the norm is for success on first flight. And even when they do fail, most of the time it is a partial failure not a complete RUD – e.g. Delta IV Heavy's first flight reached orbit, but a lower orbit than planned; Ariane 6 reached the desired orbit, but the second stage failed to relight a second time for its deorbit burn.
What SpaceX has been doing with Starship is very novel – deliberately flying very early in development, when failure is more likely than success. This is different from traditional launch vehicle design process, in which you spend years doing extensive on-the-ground analysis, simulation and component testing, resulting in a high probability of success on the very first flight.
I think a lot of people are being misled into thinking SpaceX's way of doing things is the industry norm, when it isn't. Conversely, many other people misunderstand that SpaceX is intentionally doing something different, and are judging Starship's (expected) failures by the standards of earlier programs in which they aren't expected (failure after years of exhaustive on-the-ground development is a sign either of incompetence or really bad luck).