Twitter is actually working for me still, not like it used to and it requires more manual intervention but it's working... Posts that are "flagged with content warning" get thousands of impressions on my accounts, and I've been successful in appealing them pretty much 100% of the time anyway so long as the thumbnail image doesn't have any actual nudity in it, and wording the appeal a tone that conveys the attitude of "your AI is broken again, it Mr somebody in a tank top and shorts as being naked" something similar and accurate to the immediate visual.
Just like with any type of censorship of certain terms we've had to work around before, again just look for the most commonly hashtag variations, misspellings and substituted letters of such words used to circumvent the censorship bot (eg: it might be "t33n" or "tean" instead of "teen" of "D4D" instead of "dad", etc.)
PS - I'm not saying those are actively used replacements for those words or that thise words are even censored at all, I used those for the purpose of illustrating the concept so you can apply it to the # terms you are having difficulty with