I read Donella Meadows' Thinking in Systems years after everyone told me to, which is usually how it goes. It's a slim book about stocks, flows and feedback loops, written for ecologists and economists, and it has almost nothing to say about computers. It changed how I read every outage I've worked since.
The idea that stuck is delay in a feedback loop. When the signal you're reacting to lags behind the thing you're trying to control, you overcorrect, then overcorrect the other way, and the system oscillates. Once you've seen that shape you see it everywhere: autoscalers thrashing, a retry storm hammering a recovering database, a team that adds capacity two sprints after the load that needed it has gone.
It didn't give me a single command to type. It gave me a better question to ask, which is what's the loop here, where's the delay, and what am I actually reinforcing. For a 200-page paperback that mentions servers approximately never, that's a good trade.