But police arrest a lot of people, and they aren't experts on the law. They will get things wrong frequently. They can't be expected to make the correct call every time, but they also can't do the job if the average officer is regularly sued. So the courts came up with the "reasonable" standard. If the officer took away your Constitutional rights, but it was reasonable for them to think it'd be okay, for example if it was a fuzzy legal area or an obscure point of law (what if there's a split court on whether photography can be prohibited in subways?), then you can't sue them. In effect, you can't sue a police officer for acting like a "reasonable" police officer.
So far so good. Police can keep policing without also being lawyers, but people can still sue police officers who are clearly violating their rights. In theory.
But then in 1982 with Harlow v. Fitzgerald, the Supreme Court had an issue with Nixon's aides and whether they deserved absolute or qualified immunity, and they came up with tests. Although it wasn't the focus of the case, the decision happened to grant that qualified immunity applied to every government official couldn't be personally liable for damage unless they were violating a "clearly established" right.
So now police officers were protected by a different standard than a vague "is it reasonable" test. Now the test becomes "has it been clearly established that this is a right." This isn't awful in theory, but in practice the courts have decided that "clearly established" means "has this exact thing come up before and has a court decided that this is or is not constitutional/legal?" This leads to truly ridiculous scenarios, like a judge saying "okay, sure, there have been cases where we decided that you can't use deadly force against unarmed people fleeing by car, but we haven't decided that this is the case if the car is near a highway, so that hasn't been 'clearly established.'"
Anyway, that's the state of things. There are certainly plenty of scenarios where the previous standards of "reasonable" and "good faith" would seem perfectly legitimate. There are oodles of situations which requires a tricky multi-part test undertaken by a judge, and for such cases, a police officer shouldn't be sued for getting it a bit wrong. So the basic concept isn't fundamentally ridiculous, but the "clearly established" test for qualified immunity is.
Here's an example. In 2018, a police officer slammed an unarmed man into the ground. Look at the judge's decision: http://graphics.thomsonreuters.com/srepfiles/qualified-immun...
"In so holding, the Court notes that a court can almost always manufacture a factual distinction. For example, here, McGarry was in his kitchen, while York was in a Target parking lot. That kind of factual difference and all of the factual differences listed above should not make a difference in the qualified immunity analysis, but, using Judge -- now Justice -- Gorsuch’s test from Kerns v. Bader, they “might make a constitutional difference,” 663 F.3d at 1187 (emphasis in original), so the Court must conclude that the officer is entitled to qualified immunity. While the Court thinks that a reasonable officer should be able to discern from York that grabbing and throwing an unarmed man to the ground without warning for arguing with a police officer amounts to excessive force, Justice Gorsuch would probably think that the police officer’s Taser threat in York is a fact that might make a difference."
The judge very much wants to NOT grant qualified immunity, but he's up against a wall and is angry about it.