Most of the technical books I read leave nothing behind. A trick, maybe, then nothing. The ones that stick are almost never about the thing I do for a living, and the one I keep pressing on people is a slim primer on systems thinking written by an ecologist, not an engineer.
The idea I took from it is delay. Not latency in the wire sense, but the gap between cause and effect in any loop. You turn the tap, the bath fills, but the level you see is the result of choices made seconds ago, and if you steer by what you see now you will always overshoot. That is every flapping autoscaler I have ever debugged. That is every time I changed a config, saw no movement, changed it again, and then watched both changes land at once and make everything worse.
Once you have the word for it you spot it everywhere, and you stop reaching for the panic lever quite so fast. You learn to make a change, write down when you made it, and then wait long enough for the delay to play out before you judge whether it worked. That single habit has saved me from more self-inflicted incidents than any monitoring dashboard.
It is barely two hundred pages and it is not about computers at all, which is exactly why it works. Worth the afternoon it takes to read.