It seems, after several comments I've received, that the way MiHsC (may) work on the emdrive is easily misunderstood (including by me, so I've just rewritten this blog again!). Anyway, here is an attempt to clarify it and explain why I think you get a push consistently towards the narrow end from MiHsC and why, although new physics, it is at least perfectly self-consistent. This is based on the maths in my published paper (see the reference below).
Imagine a microwave photon bouncing from end to end of the emdrive cavity (see the diagram below). As the photon goes from the narrow end to the wide end (see the lower arrows) the number of allowed Unruh waves increases because of MiHsC (more are allowed on the right because the cavity is wider) so the photon's inertial mass increases (so the right hand arrow is thicker). This means that MiHsC has disturbed momentum conservation, which can only be satisfied by applying a leftwards force to slow the photon down (its speed is represented by the arrows' length).
As the photon bounces off the right end plate and goes leftward again (upper arrows) it looses inertial mass by MiHsC (the upper left hand arrow is thinner) so the only way to satisfy momentum conservation is again to apply a leftward force to speed the photon up (the left hand arrow is longer). In both cases, to satisfy both MiHsC-induced mass changes and the conservation of momentum, a new force must appear towards the narrow end. Since the momentum at both ends is the same, the photon pressure on the end walls cancels (see this previous post). This new thrust predicted by MiHsC agrees quite well with the emdrive data (see the reference below).
McCulloch, M.E., 2015. Can the emdrive be explained by quantised inertia? Progress in Physics, 11, 1, 78-80. Link