In debates I read about similar people, "projection" is a word mentioned repeatedly. I would also suspect "reaction formation" (known as "the lady doth protest too much" outside of psychoanalysis) to play an important role.

That means, I think there is more than merely strategically shooting at one's own phenotype to draw attention away from one's own person. If drawing attention away would be the only goal, it would make more sense to try draw attention away towards some other group, also an easy target, but not including me. For example, white male nerds could shoot at white male jocks, since it is only being white and male that is considered a bad thing in certain circles. Similarly, white male atheists could shoot at white male Christians. So there must be some additional explanation.

(Not everyone is like this. There are also people who do not shoot at their own group, but at a different group, or at least at a much larger supergroup so that their own group gets a smaller fraction of attention. For example white male non-nerds shooting at white male nerds, or rich white people putting huge emphasis on whiteness and maleness and maybe also cissexuality but never ever mentioning class privilege. (Which is rather ironic, considering that the whole privileged/oppressed framwork was stolen from Marx. Here, Marx would be an example of a rich white male shooting at rich white males.))

So I guess in a way these people are trying to shoot at themselves -- on some metaphorical level. It's like they perceive something undesirable in themselves... then use typical mind fallacy to generalize it to their whole group (because being a member of a sinful group is less painful than being a sinful individual in otherwise mostly healthy group) ... and then try to atone for their sins by attacking all the other members of their group (because it is less painful than trying to improve oneself). That is, on some level they are sincerely fighting against something they consider evil. They just completely lost control over the huge biases that govern their evil-detection mechanisms.

Here is an experimental prediction: Find a sample of über-politically-correct white men publicly shooting at their own group (not just a similar group or a huge superset). Explore their background, and the background of typical members of such group. I predict that among these online warriors you will find a higher percent than in general population of racists, rapists, etc. (Where by "racists" I don't mean scoring non-zero on an implicit association test, but like actual neo-Nazis; etc.)