I cannot see any downsides to systematically documenting instances of police brutality and demanding justice for each instance. It seems entirely just to me.
I can only speculate about what your exact argument is: I'm guessing that what you're saying is that, surely, today's police are behaving better than the police in, say, the civil rights movement era (I couldn't find any data on that but I'm willing to accept that much for the sake of argument). And I think you infer from that that police violence is downward trending and that therefore, we should just let matters run their course: any upset to the delicate improvement might plunge us in the other direction.
But your inference is incorrect. You're looking at a function over time, see that this function was (probably) higher at some point in the past than it is now, and conclude from that that the function is still trending downwards. Essentially your argument seems to be based on the entirely unjustified assumption that the function has to be monotonically decreasing.
But if you don't assume monotonicity (and you really shouldn't), then function values in the far past don't give you any information about the derivative of the function at t=now. And the derivative at t=now seems to be (ever so slightly) positive, not negative as you imply. In which case, evidently, things are not improving on their own.
You make the question uninteresting by its assumptions. You pose a hypothetical situation - a white cop killing a black man - which might be interpreted as a case of systemic racism, depending on the context of the killing. But then you impose the axiom that the hypothetical world in which the situation takes place is not racist. So the situation could not have been a symptom of systemic racism, by your definition, and any query about whether it was racist has only one answer, which directly follows from your given axiom.
So I'm left confused at why you pose the question.