If I sold a $50 gold plated brick, part of what is being bought and sold is my assertion and warranty that the private key I say I put under the sticker is really there and corresponds to the public key I say it does, which I could offer in writing. If sold to an institutional buyer, their interest may not be so much in the ability to verify the bar as correctly produced, but the fact that they have a place to point a finger and an avenue for recourse if anything were to go wrong.
Yes, that's all true, but if it's not adding anything to the security or usability, then why introduce this additional bit of counter party risk? A prudent user of your service would need to factor in the odds that you might screw up a coin as well as the likelihood of recovering the lost funds. And you would need to place some caps on the liability (what if someone loaded $10 million worth of bitcoins onto a faulty coin? would you cover that loss?).
Why introduce the risk? Because it's the product the customer wants to buy, the reason they'd want to spend $50 on something they could print themselves for free with a web browser. A rational user of my service will consider that given a strong incentive to make sure I produced a functioning piece and no incentive to not do so, the likelihood that I will do so is overwhelmingly high.
I am not worrying that someone loads $10 million on to a faulty bar, because if I'm going to offer a warranty like that, I can choose to be careful enough to make sure the item isn't faulty. It's just a piece of paper with two numbers on it, one of which is partially visible and can be independently verified. Unlike an electronic gadget or product that does something, it's not likely to fail in any manner that could be considered my fault. The only thing I had to be absolutely sure of is that what is printed on the front matches what's printed on the back and that I haven't been sloppy with copies. I can't warrant that the piece is uncrackable or that it isn't vulnerable to some sort of high-tech imaging or whatever, but I can warrant that I put in it what I said I did and that it was good at the time of delivery.
Of course, I'm more willing to offer that on the series 2 windowed items rather than the ones that are completely covered. Granted, there's maybe a 1/1000 or 1/5000 chance that despite printing everything in sequential order to keep it as simple as possible, I simply put the wrong private key inside a series 1 coin. This error could occur simply because the label and key has to be matched by sight and human error rate is never zero, a risk that is still mitigated in practice by the most likely error being a transposition and that the wrong key still has 1 BTC on it. But on the series 2 coins, that risk doesn't exist, or at least is fully within my control.
Restaurant owners have it worse. On any given day, some 16-year-old new-hire could decide to play a vicious prank without thinking through the consequences and put something in a taco that turns somebody into a vegetable and runs up $10 million in medical expenses, pain, suffering, attorney bills, negative PR, and lost business. They are exposed to this risk again and again each day they do business (though naturally they buy insurance for the part of it that they can be billed for). My exposure was limited to spending 1 day designing and printing less than 100 sheets of paper and hand-inspecting each one to make sure each sheet satisfied a list of criteria to assure me each key was safe to put money on before placing the sheets into a laser cutter and turning them into circles. The next time I need to print more, I'll probably block off another whole day to do it, just to give myself the means to know I did it right.
EDIT: here is another way to put it, while still acknowledging that the error rate is nonzero (assuming, for example, a cosmic ray could flip a bit in my printer and make it print the wrong thing - or that I could make a mistake no matter how careful I was and how much thought I put into it):
if it's a $50 product, let's assume my profit is half, and the typical clearly-my-fault major error could be a $50,000 liability. (that's not to discount that $10M is possible, but it's probably on the high end of the spectrum, and quite honestly, someone with a $10M budget for bitcoins will probably do more due diligence than just buying and loading my bar and will probably employ someone to manage them). That means I would have to sell 2000 bars to absorb the losses of one bad bar. If I think my risk of making a faulty bar is much less than 2000:1, then selling bars for $50 should be a great bet for me to make.
Meanwhile, I haven't drafted that kind of document that would constitute any such "warranty", but if I were to put a specific dollar amount on it, I'd be limiting my liability at the same time offering some sort of meaningful guarantee. Bottom line is though, I believe I can produce those bars with enough accuracy that it would be like selling life insurance to the immortal: profitable no matter what price is charged. It is a guarantee I believe I will never have to pay out on.