When 4/8chan came to raid the audacity fork I created, we also had huge problems getting rid of spammy issues.
There's only a feature in github to forbid issues being created by "new accounts" which means if the account is older than 48 hours, it can still spam the shit out of your organization. And good luck trying to keep up with hundreds of accounts simultaneously spamming comments with racist and harrassment content. And this feature only exists for organizations, not for public repos owned by a person.
The only alternative is to make the github orga only accessible to triage members or maintainers, which would kind of defeat the point of issue tracking and bug reports.
Things would be so much easier if GitHub/GitLab/BitBucket had an automoderation bot that does this for you, even a simple blocklist of words or syllables would decrease most of the spam.
As most of the chans/spammers use lurking accounts that they reactivate for that purpose, they're very easy to identify. Their email addresses used in git are always from the same providers, and they always use the same kind of language, same kind of avatars, same kind of grammar and questions avoiding the main topic. So it's easy to spot them via a simple ML trained model (bert or tacotron models could easily classify this). Probably even a seq2seq model would be enough.
But yeah, guess this was just my two cents that GitHub has a moderation problem.