I read a piece somewhere that really stuck with me. I can't find it again, but the idea was that for complex systems, logging discrete events makes little sense

I read a piece somewhere that really stuck with me. I can't find it again, but the idea was that for complex systems, logging discrete events makes little sense. It's not sensible to operate complex systems by discrete events, in part because there are so many things happening, but also because it might be hard to figure out what nth-order effects events have. Instead, to figure out how well a complex system is running, you need to monitor continuous values. Things like flow volume across boundaries, level or volume in buffers or storage, composition of things as fractions of the whole, and so on. Other industries (nuclear, aeronautics, oil and gas, electricity) know this already, but us software engineers have been slow to pick it up. That resonates entirely with me, but I still have trouble putting it into practise. Does anyone have references to stuff written about this?