"those cracks haven't appeared in testing" - this is the testing. You don't just simulate the stresses, you have got a bunch of devices that have experienced atmospheric reentry in different conditions, and now you could do the tests on them.
True, but if this additional testing reveals issues not revealed earlier, they'll just iterate again until they won't have to X-ray every plate after every flight, because that is clearly not the MO they have in mind. In the worst case scenario, they might need one extra vehicle generation (BFR?) to actually get there, but I'm quite certain they will eventually get there. Otherwise there's just no point to that company.