I agree that the assumptions they made about peak forces were high. A 12kn force on the climber's side would flirt with harness failure, and would definitely injure the person. 20kn on the first piece is, as you guys have already said, too high for a real world scenario when there is ANY rope out at all. I think it's probably even high for a ff2 on a belay anchor in a climbing scenario. Rope slippage still counts there. Static lanyards are a different story, but it still seems unlikely.
The manner of the testing is what interested me. Data can be represented in all kinds of ways, but the implement they were using shouldn't have much inherent error. Most destructive tests are based on failure with a limited number of force applications. This test just confirmed that carabiners can take a lot and keep on ticking. Which we knew already.
I think it's great that people continue to TRY and mimic real world scenarios in their testing because that is, after all, where we use the stuff. This wasn't the be all and end all in cyclical loading studies, but it was an interesting experiment nonetheless.
nzcaver, I have also retired groovey carabiners, mostly because of the obvious reduction in mass precisely at the bends, where we know the maximum stress to be. In my mind I know that the overall strength loss would have been minimal, but they just LOOK scary. I'm going to continue placing those ultralight wiregates at the top pieces during a climb and not worry about cyclical loading, but mostly because they don't get loaded that often. Like you said, they're likely good for the rest of my climbing career and beyond.
pseudofission, I'll be anxiously awaiting those dynamometer tests.......