Bottke, Vokrouhlický and Nesvorný argued in Nature that a 170-kilometer asteroid broke up about 160 million years ago, that its fragments drifted into Earth-crossing resonances, and that the probability one of them was the Chicxulub impactor exceeded ninety percent, with a mineralogy matching the presumed CM-type of the boundary sediments.
In 2011, Masiero and colleagues used WISE infrared data to re-date the family: the breakup happened only about 80 million years ago, leaving far too little time for fragments to reach Earth. NASA's release kept the file open with the line that the demise of the dinosaurs remains in the cold case files. Independently, spectroscopy showed 298 Baptistina is mineralogically LL-like, not the expected carbonaceous type.
The claim was not fraud or folly; it was the best model available, published at high confidence, and falsified by a better instrument within four years. That arc, conviction to acquittal on improved data, is the rarest thing in popular science coverage, and the most useful one.
Ninety percent is as close to certain as science gets in public; what happens to everything else quietly riding on that number?