Carroll's final point – that the researchers measured thrust not only when the drive was configured to produce it, but also when set up to do nothing at all – may be the most important takeaway of all. It's a point Mika McKinnon expands upon in her explanation below.
It's funny that everyone I've read thinks that is important, that it points at an experimental error. Bull.
This is what's important. From what little I understand of it, he's right.
The business about "quantum vacuum virtual plasma" (the physics of which they "won't address" in this paper) is complete bullshit.
It is important to understand that the device in question was the Cannae design, based on their in-house theory, designed and built not by NASA, but by Cannae, LTD (out of Pennsylvania, IIRC.)
Now, if you design one thing to work one way and another thing almost like it to work another way, or not work, using an idea of what makes it work in the first place that is completely wrong - then why is it surprising that you did not get the results you expected? What that proves to me is this:
The business about "quantum vacuum virtual plasma" is complete bullshit.
There is another mechanism in play but the error shown here is an error in theory, his understanding of what makes it work or not work is wrong. But we've known that we don't understand what makes the EM Drive work every since the first information on the EM Drive was published. Just because Cannae still does not understand it DOES NOT make the cause of the measured result an experimental error.
Don't misunderstand me, this lack a theory of operation does not rule out experimental error, but neither does it add one iota of evidence for experimental error.