I have been asked how I can justify the Hubble-scale Casimir effect (HsCE) in MiHsC since there are unlikely to be conducting plates situated at the Hubble edge. So here are the two answers I normally give to that, the first when in cautious mode, the second when I indulge myself.
First: There's the old empirical way of saying 'if a simple model predicts well, then one should just accept it as being useful, and avoid making hypotheses when there are not enough data to decide between them'. This attitude has a good pedigree, Newton used it for his gravity theory and said: 'Hypotheses non fingo' (I don't make hypotheses). He meant that he didn't know exactly how gravity worked, but he could certainly predict it and that was enough. So in the case of MiHsC, assuming a HsCe allows you to predict things better, so whatever is really going on, it looks like a HsCe. Having said that, it's difficult to think about something for so long without trying to dive a little deeper..
Second: The best model I have thought of so far considers information rather than objects (appropriate in this new digital age). If you assume that the Hubble horizon is an information boundary then it's only right to go all the way, and not only should the horizon not allow information to pass through, but it should also disallow patterns within the cosmos that would allow us to infer what lies beyond the horizon. This means you can't have a pattern (eg: an Unruh wave) that doesn't fit exactly or that doesn't 'close' at the Hubble horizon, because if you did allow a partial pattern you could infer the rest of the pattern and therefore some of what lies outside the horizon, which would defeat the purpose of having a horizon. This 'horizon wave censorship' model is equivalent to the Hubble-scale Casimir effect that Unruh waves are subject to in MiHsC but can also be applied to any pattern, and therefore can also be used to explain the low-l CMB (Cosmic Microwave Background) anomaly (the observed suppression of CMB patterns on large scales). I discuss all this briefly here: http://www.mdpi.com/2075-4434/2/1/81