Yes, GR is almost without peer as being a theory that was born largely out of a philosophy of relativism, but note it could probably never have been obtained without the observations that propelled special relativity. In other words, one should ask if relatiavity was propelled by the idea that the laws of physics should be the same in all frames for some philosophical reason, or if someone just started with a "what if" and figured out the ramifications that could be tested. That's where it differs from some of the more modern stuff that the philosophical approach inspired.

Then there is also the gravity/inertia coincidence of Newtonian gravity that benefits from the unification that comes from elevating the equivalence to a principle. Unification is always a valid goal of science, as long as the price of unifying one concept is not the introduction of a more complex scaffolding of new and unconstrained concepts (the current problem with string theory). But as you point out, general relativity made all kinds of predictions that were addressible with easily imagined experiments, like the bending of starlight. The famous eclipse observation would never have been undertaken without GR, so the theory did its job by connecting previous observations (Galileo's gravity experiments, the Michelson-Morley experiment, Mercury) to new ones (bent starlight, etc.). That's really the job of good theory, to unify past and future experiments into a single descriptive framework. I'm not seeing much of that with the newer stuff, we may be running into a fundamental limit of what our technology can provide. Or we just need to ask smarter questions that we can get access to with current technology. Or, physics might be close to done, after all. Time will tell, but string theory and multiverses had better make some testable predictions soon or the effort will be largely a waste, as science anyway.