I agree that hammering random things isn't where you should start. I do sadly have to admit that "just do something" is unreasonably effective. Which means if you have other things lined up, having teams that are somewhat randomly attacking things is probably going to see more success than you'd expect.

This didn't cause an outage, but I was endlessly amused about regression test failures caused by synthetic data when the Navy's maritime domain awareness switched over to a tip file format that embedded pngs of each ship detection into an xml file and had some limit I'll never remember on the maximum number of such tips it could fit into a single file.

It was actually a testament to the thoroughness with which the NRO regression tests its shit, but the complaint was that the simulated scan the program was running detections on crapped out after something like 30,000 detections, which is already more ships than the US Pacific Fleet and probably more ships than any foreign Navy has in every ocean on the planet, certainly more than will ever be contained in a single image.

But in case I'm being 1981 Bill Gates level naive here, they are in fact prepared for the eventuality of gigantic swarms of microships taking over the oceans because of that regression test.