A trust score could solve this problem. Whenever someone flags a thread, the thread either gets globally flagged or not. If it automatically gets globally flag

A trust score could solve this problem. Whenever someone flags a thread, the thread either gets globally flagged or not. If it automatically gets globally flagged, a mod may un-flag it. Or, it may not be automatically flagged, and a mod will flag it. We can then find the probability that a user's flag will result in a "permanent, global flag" by looking at the user's history. We ought to use the lower bound of the Wilson score interval for this. This is the user's trust score, which needs to remain hidden from them. We then set some threshold under which a user's flag will be shadow-ignored. This way, we only act on feedback that has a high probability of fitting the same pattern of mod behavior. Just make sure you have quality mods who consistently and correctly flag threads.