As you perhaps know from reading my own writing on somewhat similar topics, I pretty much agree with you on most everything you've written. (Good review, by the way!)

But the quote above is just a different way to talk about what Popper called verisimilitude, which he didn't manage to formulate in a formally workable way, but which was the basic idea that a hypothesis that has passed many good attempts at falsification is robust and fit for use in a way that a hypothesis that is untested (or which has failed many tests) has not.

Popper, to my mind, was basically correct about how to do science. We do need to hold hypotheses tentatively. If you can't conclude on the basis of evidence that some idea is wrong (i.e. "falsified"), then you're probably not doing science. And anything that is equivalent to that probably is.

That there are more efficient ways to do it, that his attempts to formalize verisimilitude were not terribly successful, that even the idea of falsification is flawed (because we don't actually totally rule out a hypothesis H on the basis of a "falsification" because that would be verifying the hypothesis !H, which we can't do), and that in many areas goodness-of-fit-of-model is far more useful than any sort of pass-fail scheme--none of that really matters.

To use a mathematical analogy, Popper basically says, "Hey, this is an optimization problem. Let's use gradient descent!" and after fifty years some people are all "NO, that's so slow, we use Newton's method at least" and others say "Simulated annealing avoids local minima SO much better!". But it was an optimization problem. And the simple way to approach it was workable a lot of the time.

I like your quote. But torturing hypotheses with experiments is fundamentally what Popper was about, so I think Popper's still pretty on-target.