Imagine the oracle tells the utilitarian that almost all of the actions that we ordinarily take to be right have bad consequences and similarly for wrong actions. The cost... [would] be too high to remain a utilitarian. To remain utilitarian you have to abandon more or less every moral intuition.
This is quite mistaken. Firstly, I take my moral intuitions to concern "ordinary" situations, where "all else is equal". I think it's fine to wear mismatched socks because I expect such a choice to be harmless. But if the oracle tells us that every time you wear mismatched socks this causes a kitten to suffer agonizing pain, well, it would be bizarre to just stick to one's prior intuition as though we thought that acts of this type must be permissible no matter what. I trust that nobody (sane) has that intuition.
More broadly, it's important to see that our fundamental normative principles don't depend in any way on how things actually turn out. For the fundamental principles can all be stated in conditional terms: in circumstance Ci, you ought to do Ai. A moral theory (e.g. utilitarianism) may be represented as an exhaustive list of such conditionals, covering every possible circumstance (completely described). So if you accept utilitarianism, part of what you're accepting is the claim that if wearing mismatched socks would cause great suffering on net, in that case you ought not to wear mismatched socks. If you don't accept that such an act would be wrong in such circumstances, then you shouldn't accept utilitarianism in the first place. It makes no difference whether the circumstance is actual or not. (That's why ethicists can get away with trading in thought experiments, rather than tediously waiting around for an actual run-away trolley to come along!)
Another way to make the point is this. Our credence in a moral theory (e.g. utilitarianism) should be independent of our credences in various ordinary non-moral facts [modulo complications regarding the higher-order evidence provided by expert testimony, etc.]. So for each possible circumstance Ci, one's credences should assign the conditional probability P(Utilitarianism | Ci) = P(Utilitarianism). Whatever circumstances happen to be actual, your confidence in the pure conditionals offered by a moral theory should not be affected. Moral theorizing is, in this sense, autonomous, rather than hostage to the concrete facts about how things are.
Not all philosophical theories are autonomous from the concrete facts in this way. In particular, my previous post argued that Lewisian Modal Realism is not. The Lewisian assigns P(LMR) > P(LMR | only one concrete world). Indeed, they are likely to assign the latter near-zero probability. That's okay, because they think the condition of there being 'only one concrete world' is not a genuine metaphysical possibility that their theory needs to be compatible with (their view instead presupposes a plenitude of worlds). But if they somehow learnt that this supposed impossibility was actual after all, they would respond by rejecting their theory -- that's just what the low conditional probability indicates.
Anyway, see my previous post for more on that. For now, I'm more interested in whether anyone remains unconvinced about the autonomy of the normative. Can anyone justify the claim that the actuality of a world w (e.g. where wearing mismatched socks causes great suffering on net) would somehow undermine utilitarianism to a greater degree than the mere possibility of w?