At first, I was bewildered by the fact that they were making motorized fuel tanks fly and land back and calling it a success, before they even had a smaller scale prototype, but it seems to have gone to the opposite extreme nowadays.
Many things in engineering cannot be scaled down and still represent the physics of the full-scale system. As for going to even larger rockets - the
basic rocket equation tells us the rocket with the lowest dry mass ratio can deliver more dV, and the bigger the rocket, the lower the dry mass ratio becomes due to the area/volume scaling. SpaceX isn't just winging it and blowing up large rockets for the lolz - physics and engineering dictates their activities. A 50% scale Starship/Superheavy stack would not behave like the full-scale stack at all in many critical ways, from mass ratio, vibration, stresses, engine interaction, etc. Even though it could look identical to the full scale system, it physically would be a very different rocket, and it would tell you very little about what the full scale rocket would do.
There are a few issues I can see with that. Such big tasks really benefit from taking it step by step. By not having the hindsight of all the roads not taken, such as smaller starships launched by Falcon Heavy and/or a suborbital Starship testing at least the heat shielding, they risk building the thing with built-in flaws that will come back to bite them later. And, especially since it's been designated as the single 'next big thing' at SpaceX, there are only so many failures it can take before investors get spooked, especially given Elon's recent antics.
New Origin made a motto out of this philosophy - Gradatim Ferociter (Step by step, ferociously). Sounds nice, but after many years all they have been able to do is fly a small suborbital tourist ride. I hear they have orbital ambitions, but who knows when they'll get to that point. And, despite going step by step, they got blindsided by a failure when they considered their rocket understood well enough to carry passengers. They were fortunate that it did not fail with passengers on board, but they got lucky. I'd argue that New Shepherd, being a relatively small suborbital rocket, would have been much more amenable to repeated flight testing as even the cost of a vehicle loss could be compensated partially with knowledge of sources of failure that could then be addressed.
SpaceX has clearly communicated that losing rockets is a possibility in flight testing, and that is not a tragedy if they can learn from those failures and losses. This methodology has lead to an active fleet of Falcon 9s, a completely new operational paradigm in rocketry, many with dozens of flights under their belts. They have managed to get the full Starship/Superheavy stack through max-Q and hot staging with a full burn of 31 Raptors, and they got Starship suborbital and got a lot of flight data during that flight and re-entry. This design methodology allows them to make huge improvements flight-to-flight because they have actual flight data to guide their engineering decision-making.
The advantages of full-stack testing were recognized back in the Apollo program - the sooner they could get something looking as close as possible to a real flight article flying something looking like a real mission profile, the sooner they would understand how the vehicle would behave in the actual mission. The best way to learn to do a thing is to keep trying to do the thing, learning from your failures, until you do the thing.