One small but critical addendum I'd add is that it's less about, "The people who make magic boxes of light say they used physics" and more about "EVERY person who makes magic boxes of light, all in unison, say they used the exact same physics."
There's a huge power in knowledge consensus. It's got all kinds of flaws, but it meanders generally in a growth direction, and that's better than any other prediction system I'm familiar with.
I would argue that if two people brought me magic boxes of light saying they used physics, and they each told me about two vastly different "physics" involved with creating those magic boxes, I would instead introduce them to each other in the hopes that they might reconcile their understanding of the fundamental concepts involved and make an even cooler magic box!
What you instead want to look for is:
"MANY people who are familiar with magic boxes tried to make this guy's magic box and couldn't replicate his method."
You are looking for valid contradiction to stated facts to invalidate claims, not for a widespread consensus on the claims.
This is important because, as a single example, two people can use different methods to achieve the same outcome based on their own respective theories that have minor flaws but present strong evidence that they are in their entirety true. It is only by reconciling & finding the contradictions between the two that you can come out with a single VALID theory.
If I were a scientist I would probably have some examples of this occurring in the past, but off the top of my head I would suggest it is improbably that there has never been a situation where two competing theories that each had strong evidence were both eliminated in favor of a single unified theory that reconciled differences between them.
Nah, science happens where the disagreement grows. If there's consensus, that's not science, it's technology. That's why you guys keep talking about light boxes. Try asking 10 scientists about gravity, for example. You'll get 12 different answers. And that's not an off-by-one error.
There's a huge power in knowledge consensus. It's got all kinds of flaws, but it meanders generally in a growth direction, and that's better than any other prediction system I'm familiar with.