The Lambda-CDM model ("standard cosmological model") states that the majority of matter in the universe is non-baryonic dark matter. It's generally found around the same locations as ordinary matter, and has only very weak interaction with ordinary matter (like neutrinoes do). It sounds a lot like ether except without the relativity problems. Maybe the EmDrive can somehow interact with non-baryonic dark matter and accelerate it.
I don't actually think this is true, because measurement error sounds more likely, but unlike conservation of momentum violation it would require only a minor tweak to known physics.
Agreed. When I make hard-ish science fiction settings that feature better drive technology than the real world, I assume that either:
1. A process can convert normal matter to dark matter and at the same time accelerate that dark matter to some absurd speed -- so you can take a small reaction mass with you and use it very efficiently, but you don't have an ultra-efficient weapon system.
2. A process can push against some background medium of dark matter.
But. If there is dark matter in the vicinity of Earth (and while I agree it's "in the same places as ordinary matter," that's at a macroscopic, galactic level), it's unlikely to be at some convenient rest-y-from-human-perspective frame of reference. Which would suggest that the EM drive would work better facing one direction than others. Which is, I think, not congruent with the experimental evidence in front of us.
Local dark matter energy-density is very small, around 0.01 times the mass of the sun per cubic parsec or in SI units about 10^-16 kg / m^3. In concordance cosmology CDM neither dissipates nor collides, so it cannot easily clump into warmer (and brighter) dense structures, so it is always diffuse outside of black hole event horizons.
If we accept for the moment the existence of a pseudo-CDM identical to the concordance CDM except that it can interact with EmDrive microwaves, then the first question must be: why doesn't it interact with astrophysical masers in GMCs, affecting the structure evolution therein? (e.g., if astrophysical microwave hotspots move pseudo-CDM then there will be a bias in the location and age of young stellar objects and H II regions due to the change in energy-density). Some astrophysical megamasers have luminosities on the order of a thousand suns -- that's a lot of watts to shuffle dark matter around in your idea about the EmDrive, and that shuffling will drag dust and gas along with it.
Tweaking CDM into pseudo-CDM without it collapsing gravitationally by radiating photons directly or by colliding with charged particles is probably impossible for any but a tiny fraction of the overall mass-density of the dark matter sector. Collision or direct dissipation removes angular momentum from the particles such that they can fall a little closer to the centre of the galaxy. On a galactic scale, that leads to a much higher total energy density at galactic core at even fairly small fractions of the local energy density.
Avoiding that would require extremely demanding work and likely would have you adding a lot more new particles to the standard model to capture all the additional (broken) symmetries -- and it gets worse the greater the local energy density or equivalently the greater the interaction cross-section you require in order to interact with a physically small device like an EmDrive.
I don't actually think this is true, because measurement error sounds more likely, but unlike conservation of momentum violation it would require only a minor tweak to known physics.