Two important points to keep in mind when evaluating the dueling expert reports:
1) This is a civil case, where, to prevail, the plaintiffs must simply prove their case by a "preponderance of the evidence." The labels' burden is only to prove that their version of the facts is more likely to be true than Thomas'; even 51% will do. (This is not a criminal trial, where guilt must be proved by the much higher "beyond a reasonable doubt" standard.) Thus, in evaluating whether Thomas uploaded and downloaded song files, they question is not: "Is there any way to imagine a scenario under which Thomas did not do it?" The question, rather, is: "Is it more likely than not that she did it"? So it would be perfectly proper for a juror to hear all the evidence and reason: "I thought the plaintiffs' evidence showed that she probably did it. The defense did raise some good points, and their expert convinced me that there were ways that, at least in theory, she might not have done it. But, on balance, I think there's about a 70% [or 51%] chance that she did it. Therefore, I find for the plaintiffs."
2) The labels' case does not rest entirely on a link between the IP address detected by MediaSentry and Thomas' ISP's determination that that address was registered to her. That is certainly one piece of evidence -- and it's an important one -- but there is much more. For example, the KaZaA user ID detected was "tereastarr@KaZaA" -- and, as Ars correctly points out, "Thomas was...shown in court [in her first trial] to have used the 'tereastarr' name for various personal Internet accounts." There is also additional computer forensic evidence relating to a MAC address and the examination of the files on Thomas' hard drive. And -- and this is crucial -- the jurors in the first trial listened to Thomas testify that some unknown third party must have wirelessly hijacked her ISP account, and those jurors simply did not believe her. (As Ars dryly notes, "Thomas apparently did not use a wireless router, however.") As one juror told Wired after the first trial (in which they found for the labels and ordered Thomas to pay damages of $222,000): "Spoofing? We're thinking, 'Oh my God, you got to be kidding.'.... She's a liar." The importance of the defendant's demeanor and other indicia of credibility on the witness stand can hardly be over-stated.
The retrial (required because the court determined that it had issued an erroneous jury instruction in the first) is set for June 15. But don't be shocked if it settles before that.
(Disclosure: I worked on an amicus brief in opposition to Thomas' motion for a new trial.)