All we can do is try to make a society where that doesn’t matter.
IMO such a society is as anarchic as possible, though “as possible” is still well short of the level in, say, The Culture; the exact level is dynamic and tech-dependent.
I mean, we can; we just don't. It's not like there's something baked into the laws of math that says your society is required to be a surveillance state (unlike encryption, where the laws of math do say this is always possible).
It is absolutely within the realm of technological possibility to build a society with largely decentralized infrastructure that doesn't constantly phone home to report on you to the Great Eye. We don't live in that world because normal people are kinda retarded. In the words of the creator of the Great Eye itself: "They trust me. Dumb fucks."
Indeed we could, as you say, construct societies without that capacity — Amish, etc. already do so — but such a society is outcompeted by every society which embraces tech, and any society with tech at the level of the Stasi (i.e. both old and the wrong side of the Iron Curtain) can surveil whoever it wants whenever it wants.
Now? Now it doesn’t matter if you decentralised all the infrastructure, the tech is too cheap to avoid total surveillance.
Now, laser mics are school projects, and the hardware cost for pointing one at each and every window in London 24/7 is significantly lower than the annual cost of the Metropolitan Police Service in the same city.
Now, your WiFi can be converted into a wall-penetrating radar, do pose detection, heart rate and breathing detection.
Now, my wristwatch knows when I walk past the charging station to turn on its screen and remind me of its existence. I don’t even know how it knows when I’m walking past.
Now, I have an IR camera that can see through some opaque-to-visible-light materials for no good reason and at pocket-money prices.
“Centralised” has its problems, but getting rid of centralisation isn’t enough.
Today, it is my impression that much data is only used to reconstruct events after the fact, rather than gaining a priori knowledge to prevent an incident in the first place.
(And, to make it clear - I am not suggesting 'progress' as outlined above is desirable...)
For example in France we're in a permanent "state of emergency" since the attacks in 2015 (and now with covid), which grants more rights to the government/police and let them bypass some legal safeguards for "the greater good" but of course it's already being abused, not against terrorists, but against protesters, people squatting land to protest against projects that would have a negative impact on the environment (https://en.wikipedia.org/wiki/Zone_to_Defend), &c.
But I'm wondering if that really is what needs to occur before we consider a a state to be a complete surveillance state, if, for example, almost every home has some kind of device that can be exploited for surveillance, e.g landline, mobile phone, television, smart devices, laptops/desktops/tablets, internet routers/modems, smart meters, smart heating, rubbish collectoin etc?
I.e. have we opted in to in-home surveillance to a sufficient level to make mandatory surveillance almost of no use.
Ratcheting up automation is what would make it different to me. While there are exceptions, for the most part, police look at all this data after they know about a crime.
They have most of what they need to use the info to discover crime and automatically cite people. As a simple example, ANPR in two places could issue a speeding ticket if the time elasped between two scans is low enough. I'm sure it's happening somewhere, maybe a toll road. The difference would be when it's widespread.