I think the crux comes down to this section of the exchange.
> Camgunz: points out that incarceration is expensive and this could support a lot of social services.
> Natsu: questions again how social programs will stop violent crimes like rape when existing programs have failed.
How __I__ read this (and I'm guessing Camgunz) is that camgunz provides context to ticmasta's questions about being economically viability by noting how much we spend within the prison system. Implying that this is the metric that we are working with. There was an assumption made here, which is extremely common, that policy just adds cost and does not reduce it somewhere else. What Camgunz is suggesting is that by spending money elsewhere we can reduce the costs mentioned. IMO a net even would be a gain in the sense that we are being more humane and maximizing individual freedom (considering being in jail means pretty low individual freedom as well as potential long term consequences like being unable to vote in the future).
Natsu then made a non-sequitor argument that is commonly associated with dog whistling (it may not have been intended this way, but that is part of the power of dog whistling. By definition it is coded language). The supposition that Natsu implied was that rapists are going to rape and this cannot be prevented, therefore we should just contain them. The data does not support this nor is it effective because we can only prosecute those who have already committed an action or are in the act (this isn't Minority Report, we don't do pre-crime and I think we'd all agree that is highly immoral). Further Natsu said that prisons already do not solve the problem, therefore we should just contain them. The context of the discussion was about reform, which acknowledges the fact that prisons do not accomplish this. Starting with mabbo we get a claim that we should be focusing on crime prevention, which would include (presumably) changing how prisons work. With a focus on reformation as opposed to punishment. Additionally, Natsu's statement is associated with an "all or nothing" approach, which is not what anyone here is suggesting and is a common technique used to derail a conversation, because there will always be outliers in data (i.e. letting perfection get in the way of improvements). Simply, the discussion is about reduction not abolition. Lastly, the jump to rape is one to extreme cases. A drive to an emotional appeal above everything else.
Simply, the response was not in good faith and had a high potential to be intended to derail the conversation. It exhibits many of the tactics that people use to perform such a task (again, it resembles dog whistling enough that it is reasonable to be taken that way).
I will also direct you to my response to Natsu where I link evidence of a negative cost policy (legalizing and taxing prostitution) that fulfills said objectives. I also stress the focus on reduction over abolition. There are many other policies similar to what I mentioned, including reform around marijuana laws (referencing back to Camgunz, which would you rather have? Spending 80k/yr for a non-violent pot dealer (max punishment is several years) or taxing pot, regulating to increase safety, de-stigmatizing and allowing people to seek professional help if it is negatively impacting their lives, and reducing funds that go to cartels that also contribute to migrants fleeing their countries and heading to America? The former costs a significant amount of money while the latter distributes money through the economy and allows taxes to be taken off the top). Not all policies will end up working like this, but again, this isn't an "all or nothing" situation, it is about improvement through iteration and utilizing the data that we have to make better policy. These two are low hanging fruit but do illustrate the point being discussed here.