Let me suggest the following way of looking at it. What evidence do we have suggesting that the EmDrive works? We certainly don't have a device observed to be

Let me suggest the following way of looking at it. What evidence do we have suggesting that the EmDrive works? We certainly don't have a device observed to be generating large thrusts reactionlessly. What we allegedly do have is a bunch of measurements that, on the face of it, suggest that maybe we could make one by refining this technology. But the path from those measurements to that conclusion goes through a load of physics, and nothing on that path is known with any more certainty than conservation of momentum is known under such "ordinary" circumstances. Suppose someone comes to you with a purported mathematical proof that 2+2 doesn't equal 4. You look at their proof and don't find anything wrong with it. Are you justified in believing that 2+2 isn't 4? No, however carefully you check the proof, because the axioms and rules of inference they use in the proof are going to be no more certain than 2+2=4. If you check the proof carefully enough then you're entitled to conclude that something is very wrong in our conventional understanding of mathematics, but you can't tell what and you should be pretty confident it isn't 2+2=4 until someone demonstrates how every time you put 2 things and 2 things together you're really getting 5 things and just not noticing one of them. Similarly, if someone comes to you with a purported demonstration of momentum nonconservation in "ordinary" circumstances, then if you check their work carefully enough you may be entitled to conclude that something's wrong with our understanding of physics. But you shouldn't leap from there to "momentum isn't conserved" until you actually see momentum clearly not being conserved, or find a coherent theoretical framework that explains what you've seen along with all the previous evidence for momentum conservation, and that has momentum not being conserved as claimed.