And vastly overestimating the quality of the surveillance cameras.
Ultimately, the big pushback has to be against normalization of use of special measures that are devised for special circumstances in general cases.
All the server gets is a set of random numbers from a user, after that user tests positive. And while it might be difficult to keep the server from knowing which user tested positive, it's not even necessary to store that.
The majority of CCTV is privately owned. The vast majority of CCTV is not proactively monitored. Tracking someone using London CCTV is someone staring at a screen for a number of hours trying to follow them on playback.
I’ve had the joy of recreating someone’s movements in a nightclub from CCTV. A 6 hour night out took me days to reconstruct, and that’s using a small number of cameras in a well defined area.
CCTV as a mass surveillance tool just doesn’t work.
It is until it isn't. Soon it will be a ML system and it will be possible, then what?
Then it's good that in the Apple/Google proposal, that information isn't broadcast. The bluetooth transmits random numbers. Your phone stores the numbers it has sent and received. Upon positive test, the numbers are published and all the phones compare them to the numbers they've received; people who detect a match have an incentive to get tested, so there's no need for a central authority to know who the contacts were.
I don't see how this adds much to government surveillance capabilities. Certainly not in the U.S. where detailed location histories are available for purchase. (In fact, a case in 2018 put specific constitutional limits on the government's right to purchase that data, after the government used it to track down illegal immigrants.)
Given the vast number of surveillance cameras I saw last time in London, I doubt this would add significantly to their capabilities either.