In the short term, demand for violence is effectively fixed, so decreases in the supply of simulated violence lead to increases in actual violence as a substitution effect.

This is undoubtedly possible, though I'd expect far less of a substitution effect than you because of the distraction effects I suggested above. Ultimately I suppose this is an empirical issue.

In the long term, exposure to violence leads to desensitization, so demand for simulated violence expands to meet the supply.

I suspect that once the level of simulated violence in a real society is above some saturation point, further increases in its supply would not be met by increased demand. Ideally there'd be some way to empirically test this too.

Given two otherwise-identical societies, in which one strictly limits the supply of violent imagery and the other does not, I predict that the former will (eventually, due to desensitization) have a higher demand for violence, leading to more actual, physical violence during blackouts.

I smell a Freakonomics chapter!

Seriously, if there are any economists or sociologists reading this comment, I think something like this could make a cute topic for a paper. Some quick googling makes me think that the effect of blackouts in general on crime hasn't been researched rigorously - I'm mostly seeing offhand claims like 'looting during blackouts blah blah blah' or studies of individual blackouts like New York '77. I see even less about using blackouts to assess the effect of violent media specifically, but I'd be very interested in the results of such a study.

At any rate, your own prediction is an interesting one, if only in terms of thinking about how one could test it, or approximate testing it.

As for which variations on the Langford basilisk I'd be OK with banning: I'd work it out by putting on my utilitarian hat on and plugging in numbers.

This isn't a fully-general argument for censorship of any given subject that provokes disgust; it's quite specific to violent pornography.

More than that; it's specific to media that (1) desensitizes some viewers and (2) have actual violence as a substitute good, which arguably includes violent non-porn as well as violent porn.