We believe the paper with the most authors ever recorded (a massive 2512!) is Aleph et al. Precision electroweak measurements on the Z resonance. Physics Reports 2006, 427:257–454 (available at http://arxiv.org/PS_cache/hep-ex/pdf/0509/0509008v3.pdf ) — unless somebody knows better...?
Lutz Bornmann and colleagues have tried to find out how much editors look for signs of research misconduct when assessing manuscripts. They reviewed 46 studies that reported editors’ and reviewers’ criteria for judging papers but found that none of the main criteria listed was related to detecting data falsification or fabrication. The paper is available at Scientometrics 2008, 77:415–32. doi: 10.1007/s11192-007-1950-2