You design to the needs. I have resilience in my home DNS so I have a 100% availability bounded on the home (if the home is burnt down then I'll lose DNS)

At work I designed DNS to survive the loss of a city, so again 100% uptime. The riskiest part is deploying changes, which is why the deployment doesn't run in parallel and confirms state before deployment.

Resilience isn't hard when you aren't trying to serve a billion customers. The problem is we are addicted to a "winner takes all" economy, which leads to fragility. If Tesco breaks, I can go to Sainsburys, or Aldi, or Lidl, or Iceland, or get in the car and go to Asda or Morrisons. That's diversity.

But in IT the goal is all those follow the same gartner reports and all use the "best supplier" for their backend infrastructure, and if that supplier breaks then they can't restock any supermarkets.

That's a terrible risk and needs to be stopped, not encouraged. But the individual incentives are for the CTO of each supermarket to choose what the consultants tell them. If they don't they are accountable for the choice of not using things like "industry standard".