Instead of forcing user to set a low cost limit and missing a viral opportunity or the platform writing off the massive bill the customer can't afford... just put the billing mode into a reduced price mode or have some more nuanced configurations. Sometimes is just asking the question the right way. Instead of "max spend limit" or similar "If your site goes viral, how many requests do you want to serve before going offline? 1M=$20, 10M=$100, etc" at this point, I feel like bandwidth consumption is a bad metric for billing; just use requests/visits/actions and price for those.

This is not prescriptive just illustrative. The point is make a better pricing formula to account for these massively unexpected events. Couple it with an aggressive notification policy when this traffic event gets triggered. The user should know the traffic pattern has changed and a high traffic event is happening. They can login and change the configs and decide if they want to keep it going or not.

I agree. I also agree that when dealing with large numbers of people, there will be people who don't understand this and/or will actively try to social engineer their way out of their own decisions.

Setting customer expectations and meeting them successfully isn't easy.

The fact that most cloud operators don't have actual hard cutoffs to maintain financial responsibility is intentional. Azure does, but only for specific account types. If it's PAYG, you can't do it. The end result is if you do something "weird", or someone DDoS's you, you're liable.

With a hard limit, a DDoS just takes your site offline.