I run a service that uses a mix of automation and manual help to deal with customer service issues (throwaway account). 97% of the time, things work really well, we have great customers, we're small and bootstrapped but growing and are very happy with the organization we're building.
But for whatever reason, we've become very popular on conspiracy theory websites, and maybe 3% of our customers ultimately end up using our service in a somewhat harassing way. We've modified our terms of service to ban a variety of forms of harassment, but it's often hard to weed out the problem users from legitimate ones until things get bad, and we're trying to come up with consistent policies of weeding out bad users rather than we-know-it-when-we-see-it.
For reference, these users are likely very mentally ill. The New York Times did a piece on the segment of the Internet that has found us:
http://www.nytimes.com/2016/06/11/health/gang-stalking-targeted-individuals.html?_r=0
Anyone run into anything similar? Advice?