I relate to the preemptive mitigation angle, I wonder though who would point compute power towards nuisance spam? These accounts run no scams, like common with email, they also seem to post too infrequently for DDoS attacks. Do translate type models or algorithms have a way to mark and mute nonsensical text?