I honestly think it’s much more likely that the reason tumblr hasn’t been removing Nazi blogs is that they’re just that incompetent and out of touch with their userbase and the definition of what a functional website is. I’m not completely disregarding the idea that some of the staff are Nazi sympathizers, but incompetence seems much more likely as the direct cause. I mean, they still haven’t switched to https, and they took years to even start to try to deal with pornbots.
And in this case, the obvious hate blogs are just a symptom of a greater sitewide problem. Tumblr is definitionally memetic. It also has a fairly young userbase. Combine that with a lack of administrative moderation or competence, and use of the site as a radicalization tool happens very, very fast. You get Nazis that way, but most people who are old enough to use this site do have a basic awareness that Nazis are considered bad and that many people consider overt racism to be bad. Meanwhile, other radicalizing hate groups or ideologies fly completely under the radar. TERFs and SWERFs, for one. People don’t generally know what their ideology is beyond what is apparent in the name, so it is easy for it to spread memetically. Heck, it doesn’t even have to be entire organized groups; remember when everyone was convinced that all emotionally abusive parents had NPD (because of one post that spread a lot), or that otherkin were transphobic (because of nounself pronouns, a couple people freaking out, and general confusion about what exactly an otherkin was)? Ideas roll around in any environment, but tumblr is like an ideological tennis court where every time you hit the tennis ball, you get two identical tennis balls, and they’re all bouncing everywhere, and some of the replicating tennis balls are filled with explosives. And if there’s no moderation, there’s no bomb squad.
It’s definitely possible for a site to have too much moderation. (Writscrib, to give an example of an upcoming alternative, seems to be against both sex workers and fictional media depicting underage characters in sexual situations, the latter despite being based in the US.) However, specifically in a memetic environment, it’s probably better for there to be a slight excess of site moderation that can be negotiated down, than too little moderation, especially if what isn’t moderated is negative exhortations or hate speech. We’ve seen the result of too little.
Yes, it could also be leveraged against people doing good. However, it’s harder to create and spread a shutdown campaign when memetics is regulated, or, for that matter, if the regulators and moderators are actual people. One of tumblr’s many problems is that it keeps trying to do what little moderation it does have via algorithm. Algorithms are a lot easier to trick than humans. (It’s not just about abuse report abuse. Tumblr’s new safe mode algorithm tries to identify nudity based on the percentage of flesh-toned pixels. This has led to numerous instances of sfw pics being flagged because they showed beige furniture or carpets, and probably nudity could avoid being flagged by simply using a color filter.) Having moderation done by carefully vetted real people, with good diversity training and with a focus on harm reduction and anti-bullying rather than moral panic, would be the best idea.
Note that the level of acceptable moderation for a site is directly linked to its reliance on memetics. Tumblr? Highly memetic; most posts are reposts, not original content, and memetic spread is very fast, with memes and even discourses often becoming passe in a matter of days. Twitter is similar. AO3 or WordPress? Low memetic. WordPress has reposting capability, but the format of its posts means that it does not have rapidly spreading memes or slogans. AO3 has no reposting; the closest thing it has are the “inspired by” field and the ability to post to collections, plus its tagging system. Both allow users to moderate comments on their own posts. Tumblr, Twitter, WordPress, and AO3 all have similar levels of staff moderation—fairly low, especially in terms of moderating content (most of AO3’s moderation in particular is concerned with searchability formatting). Guess which of these sites have more problems with harassment? The memetic ones, tumblr and Twitter.
A Nazi is still a Nazi wherever they post, but their ideas will spread and be reinforced more on a high-memetic site than a low one, so letting them alone on a high-memetic site is much worse than on the low one. Advocating for more moderation of content on tumblr and advocating against more stringent moderation on AO3 is not hypocritical; it’s simply acknowledging that the situations are different and that they require different actions, and that the real harm of hateful ideology, on the large scale, is not “does someone have it,” but “is it being reinforced and how fast,” and “what capacity do they have to spread it?” Moreover, ideological memetics on tumblr and twitter, tend to be presented as philosophy that one should believe, whereas the closest thing to memetics on AO3 is erotica tropes that are specifically presented as fictional and often as things that should not happen in real life, and that is an important difference as well.
Probably, someone with more experience in sociology and/or web design should take over this topic from here, but that’s my onion.