On a large public wiki, a witch-hunt is the attempt to discover or assert that several userids or IP numbers "are the same person" (see ad hominem) and to seek to discourage them from contributing by means of asserting some negative reputation, and further, using that as an excuse for ad hominem revert or even ad hominem delete or (most seriously)outing.
As an actual behaviour-shaping tactic, this rarely or never works, as genuine trolls seek such negative reputation as a sign of their prowess - in particular they seek to be extremely hated by inquisitors, since "the enemy of my enemy is my friend" and others who object to witch-hunts on principle will eventually find and offer direct help to the troll, thus making the wiki troll-friendly. Co-operating patrolls tend to work in pairs or more, and may actively cooperate to drive witch-hunters crazy and make them make wrong decisions. This is simple, as such hunters are already foolishly attempting to do psychiatry without a license, and, already making wrong decisions almost by definition. On the Internet it is never easy to be sure of identity, with so many IP proxies and other people's text easy to retrieve and re-use.
There are many good examples, but, it would be a poor idea to list any, since, that would only encourage inquisitors to seek out and punish those who make troll-friendly offers. That pretty much says it all about their integrity and tactics - they see the world in simple terms of friends and enemies. Sadly this requires everone else to, as well, and note enemy projects that they have taken over, and which will tend to train others to employ the same tactics.
Consumerium should resist or reject such people to the point of letting them be driven off by trolls. This is not a perfect solution and should be employed only for inquisitors and the more persistent GodKing minions - those who literally can't seem to comprehend any other way to do things. It should eventually rely on factionally defined ways to seek out and assign editorial judgement to particular types of assertions, and rely on factions to approve or disapprove edits, so "who wrote this" is never an issue. In fact, that is pretty much the only way one can get identity out of the editorial decision.