I am happy to see a discussion about horizons, because this will give me the oportunity to ask some questions that bother me a long time ago...
The cosmological event horizon is actually an observer dependent horizon, same as the Rindler horizon. What bothers me, specially in case of the Unruh radiation, is the following: You can get the result of an observer dependent horizon considering only the action of some field (a scalar field for example) in flat spacetime viewed by an accelerated observer. This observer will detect a thermal bath of particles. Calculations and qualitative discussions I am aware of stop at that point. However, a thermal bath of particles should create a gravitational field that should pertub the curvature making it nonvanishing. But, does this make sense? The inertial observer and the accelerated observer, both in the "same" spacetime, would measure different curvatures.
There are other things that bother me, like the fact that everywhere it is mentioned that the thermal radiation is "emitted" by the horizon. The thermal radiation from horizons is a mathematical consequence of the Bogolyubov transformations between two different vacua. In general, these lead to particle creation when evaluating the number operators in different vacua and I would expect that this is a homogeneous and isotropic and nonlocalized phenomenon of particle creation. These particles do not form a thermal distribution in the general case and thermal distributions appear only in case of horizons. Thus I would expect the thermal case to be a special case of the general case of particle creation between two different vacua. However, everywhere one can read that "horizons radiate", meaning that they produce the thermal distribution of particles. According to the more general case I would expect the thermal radiation in case of horizons to be also a bath of particles, a kind of "diffuse" radiation, isotropic and homogeneous, but not a radiation coming from the horizon.
