Two sorts of social systems seem particularly vulnerable to individuals who act in BadFaith.
This is usually alleviated by giving some individual humans broad power to interpret or even disregard the rules; for example, in RobertsRules, the chairperson can refuse to entertain dilatory motions; in U.S. court, the judge can interpret procedure and, to a large extent, the law.
But this always gives those same individual decision-makers an "unfair" amount of power. In fact, although the human oversight makes it harder for others to game the system, it makes it easier for the overseers to game the system (for example, chairpersons in the United States Congress often use their leverage to further their agenda). So, you run into the WhoWatchesTheWatchers? problem, and have to strike a balance between the two extremes (systems with no human oversight, and systems run by human dictators unconstrained by process rules)
The problem is that the attacker can claim to be simply stating an unpopular viewpoint, and can claim that any attempt to ignore them or shut them up is at the best GroupThink and at the worst, unfair repression.
At this point, even (especially) if there is no formal FairProcess procedure, the desire of the other community members to be fair can be leveraged by the attacker to draw out an argument almost indefinitely. Each time the rest of the community attempts to move on from the argument, the attacker will often find some way of making some members of the community feel that s/he may be being treated unfairly.
When the attacker's luck runs out, s/he can simply come back as another user and start over again. Perhaps the attacker can use their various identities in concert to make things even more confusing. In real-life communities, not only would such an attacker get only one shot (one identity), but the cost of ruining their reputation is higher, so most people would not find it worth it to attack in the first place.
Ironically, in this situation perhaps a more formalized FairProcess rules could help. A formalized set of rules for FairProcess would make it clear when an individual has exhausted the "appeals process". Often the community members mostly feel that the attacker is acting in bad faith, but since no "appeals process" has been pre-specified, the attacker can continually try and claim that the ad-hoc process to which they are being subjective is inherently unfair, and that therefore they need another "appeal".
To AssumeBadFaith? is important when dealing with vigilantes and vandals. If you do not, then, they will exploit your trust by waiting for a moment when you have foolishly given them some access or power over you, and then betray you.
When someone has been touched by Sysopism they don't get better fast. Believing in one Echo Chamber lie creates more echo chamber lying, then GroupThink, then, Sysop Vigilantiism. Basically, people make themselves stupid by believing the Sysop Power Structure, which they do to have less casual conflict. Gradually they consider the SVpov to be just "Common Sense" - then they pollute Politics As Usual with those foul beliefs.