The police should show that a crime has likely been committed, and get access to just the data that probably has evidence.
There are many other contexts where we trust properly supervised people who lack an immediate and obvious incentive to abuse the system. Combined with good overall software design, auditing and transparency almost all of the harms could be mitigated. And the tech does have some pretty major benefits.
The cameras apprehend criminals. I can show with evidence that the juice isn't worth the squeeze, and in fact that the cameras had the effect of tasking our police force with doing municipal debt collection for Melrose Park and Maywood, at the cost of 5-7 hours of sworn officer time per "failure to appear warrant" arrest. But supporters of the cameras will point to multiple stolen car interdictions and recovered firearms.
If you go into these kinds of things assuming that the median resident of a municipality is anti-policing, you're already way, way off. And I find when I talk to anti-Flock advocates (that is: people who have "anti-Flock" as part of their personal identity, not just a person chosen at random who would happen to answer "no" to "should we ALPR") that many of them are operating from anti-policing premises, and so these kinds of responses are very surprising to them.
(Totally reasonable for your reaction to this to be "whoah, that was a lot more than I asked for", I just feel like I've been in these kinds of conversations a lot. It's not personal.)