[outworlder]

> We’ll never measure the interior of a black hole, for example

Don't say 'never'. You can say we are pretty confident that the interior of black holes can't be measured now or in the future, given our current understanding. But science is not made of absolutes.

> What do we do with a theory that conforms to every measurement ever made, but whose predictions aren’t testable

Then it's kinda worthless, at last for physics. If two explanations agree with all observations (so far) that means they are completely interchangeable. You might as well use Occam's Razor and pick the simplest one. Until such time we find experimental differences, then we can compare again.