I think Josh's theory (the main part, . . . he got off track a little in the last half of his post) almost accommodates everything.
I'd modify it slightly.
It could be that we have a mechanism designed to increase appetite when we sense that there's a sure source of non-poisonous food. (this is slightly different from saying that we have an aversion to "that which is not food").
And we determine that a source of food is non-poisonous through a strengthening of the flavor/calorie association.
The idea here is that if a food doesn't agree with us, we won't eat it (from disgust, or nausea-related aversion), and the association won't become strong.
If the association between flavor and calories is strong, that indicates we've been eating it happily for a while. Therefore, let's stock up! It's safe, and abundant. Let's go!
Now, if we take the oil or sugar water, it's not that it's "that which is not food" that causes AS. Instead, because there is no calorie-flavor association, the "let's stock up" trigger is not tripped.
Similarly, with crazy spicing, the flavor calorie association is not strong, because it's a novel flavor.
And with nose-clipping the strength of the signal is muted, because part of it is blocked, and so it doesn't trigger the "let's stock up" response either.
So it's more that the overeating signal is not triggered than that the "poison" alert is given.
It's more a matter of being cautious about loading up on a source of calories until after we've had several good experiences with that flavor than of going into panic mode with an unfamiliar flavor.