Imagine a ball in space. Strictly speaking in physics and especially in quantised inertia you can't start talking about it being stationary or not because it has to be moving relative to some other object, so let's say it's static relative to a nearby teapot, but far enough away that the attraction from the teapot is small.
Now put a horizon on one side of it. According to quantised inertia this will damp the Unruh waves from the direction of the horizon and so the ball will be pushed by the imbalance in the Unruh radiation field towards the horizon. Another way to think about the same thing, the informational way (see reference) is that the horizon deletes the knowledge the ball has about the cosmos beyond it. Landauer's principle says that every time you delete information, say, you erase 101011 to 000000, then entropy decreases. That cannot be allowed, so the second law of thermodynamics says that high-entropy heat energy must appear to compensate. So computers get warm when you erase data. I've calculated this energy for the deletion of space, and it turns out to be just enough to power the movement of the ball predicted by quantised inertia (see ref).
So the ball accelerates towards the horizon. Now, as pointed out by several people online or in emails, what happens if suddenly the horizon disappears so the ball gets back all its knowledge about the cosmos behind it? The problem is, it still has the kinetic energy it picked up from the loss of information. Does it lose the energy when it gets the information back? The answer is not necessarily "Yes", because although the second law of thermodynamics says that 101011->000000 must release energy, there is no such imperative for 000000->101011, since there is no drop in entropy.
Can we use this asymmetry, and repeat the process to generate energy? I think that is what is happening with the cycling photons (near and horizon, then far..etc) in the emdrive. However, this brings up many fascinating new questions to ponder. Where is the information 'stored' while the horizon is close, so the system can get it back when the horizon is gone? Can information or heat be swapped between reference frames? How does this relate to the black hole information paradox?
Getting philosophical for a moment it makes sense that our new ability to model worlds ourselves (simulations, games) is inspiring new models of the one we are in, including my recent attempt to express quantised inertia using information theory. Is it just the latest useful analogy? (Probably). Is the cosmos a self-evolved bit-system? Or are we in a deliberate simulation? I'm sure the theologists will spend many a happy hour discussing that!
McCulloch, M.E., 2020. Quantised inertia, and galaxy rotation, from information theory. AdAp (accepted). Summarised in my ANPA talk here (the relevant bit starts at 16:24)