That's an interesting take. I will share just a single example that flew by earlier this week that is clear evidence of this policy not being followed, either by algorithms, or by manual followup.
https://twitter.com/ButNotTheCity/status/1518051399631785986
Content moderation can be a challenging problem, but this is a clear failure of both algorithmic and human moderation processes, and there are an enormous number of these failures that lead to real harm in the form of radicalization and targeting of individuals and groups online and the real world.