I've sometimes seen people say that they need concrete simple examples of ideas like expected utility and Bayes' theorem. So, continuing in the same vein as An Abortion Dialogue and Case Study: Melatonin, I recently polished up my shorter-but-hopefully-still-interesting article on Console Insurance.
It's basically a short discussion of how back of the envelop estimates show console insurance (and by extension, most warranty extensions) to be a bad investment.
I write in response to your introductory paragraphs.
Interestingly, most insurance companies give out more money in claims than they take in in premiums - this being a cost of doing business that must be paid for the privilege of holding the "float", being the money held by the insurer which has not yet been paid out as claims. This money is held in some form of investment, often dominated by government bonds, which produce a yield small relative to the float - but large compared to the capital invested in the company as the float should be much greater.
Thus, for an average cost which is slightly less than the yield one would get on their invested premiums the risks of a catastrophe are sold. I would generally refer to that as "fair" as you put it, though I would prefer to avoid the pitfalls of such a subjective term, and I would agree that extended warranties are an "unfair" method of abusing consumer ignorance.