This system wouldn’t work in place of moderation, but rather alongside it. The two would have an enhancing effect on each other:
- Reach limits greatly limit troll effectiveness, since they can’t find each other as easily
- Posts that exceed the threshold naturally vs. being trolled past would have different “fingerprints” that could be used like a blacklight for troll detection for both assisting moderators and for model training for automatic suspected troll flagging
The threshold should probably be dynamic and set at the point at which posts “breach containment” (escape from their intended audience), which is where problems tend to occur.
Bluesky-like self-moderation controls would also help.