Take power plants for example: they are all interconnected in large regional grids and deal with such large quantities of power that bringing a power plant online is a dangerous process that has to be coordinated with the rest of the system. You have to spin up the power plant and make sure that you are in phase with the grid before connecting your power output - all while accounting for geographic distribution losses, predicted load changes, and so on. Theoretically, any discrepancies (like power flow being too high due to phase difference) would activate breakers and other safeties should an adversarial actor decide to bring a power plant online online willy-nilly. Unfortunately, the last few decades have shown that those safeties are barely able to handle common scenarios, let alone an attack. There is a serious risk that a single power plant could cause systemic damage to many devices in a region's grid but for the entire history of the US, we have just assumed that anyone with tens of millions of dollars to invest in a power plant is just too profit-driven to pull off anything like that.
If the last few years (decades it seems) have taught me anything, it's that we have to reevaluate many of the assumptions we have been holding about reality and our peers. From the Target/Home Depot/Equifax breaches to Facebook's Cambridge Analytica fiasco, we have just been way too lucky and way too trusting. GPS is just one of many such cases, although a surprising one. I never thought the world would put so much trust in a system created by and for the US armed forces.