For a slightly fictionalized version of why these might be useful in the real world, let's say a Finnish life insurer is selling life insurance in Hungary. So, they're doing their accounting (the numerare) in Markka (FIM), taking in payments in Forints (HUF), and exposed to changes in Hungarian mortality rates. So, for the next 30 years, this insurance company wants some secondary protection against adverse shifts in the Hungarian mortality curve, but to make that insurance less expensive, they only want protection if those losses in HUF, converted to FIM exceed 100 million FIM.
So, this insurance company goes to the structured products desk of large multi-national firm and gets a very bespoke option created. The multi-national sells this single contract to the insurance company, and hedges the individual FX components using FX options and the actuarial risk with actuarial derivatives. There's basically no market for this stuff, so the large multi-national basically commits to hedging out the other side of this contract for the next 30 years.
I've worked with optimizing risk modeling for exotics, because it turns out that for some of the more complex exotics, the daily computational costs get very expensive for basically re-calculating a first-order multidimensional Taylor approximation (the risk components) of these contracts. Every day, we then take these risk calculations and buy/sell a bunch of simple options to bring the first-order partial derivatives (risk components) back close to zero.