One of the most effective ways to manage this situation is to simply moderate the posts of all new users who do not meet some criteria (number of approved posts + some other metric, for example).
Then, it does not matter how many bogus accounts or proxies the trolls creates, all their posts are moderated. This is basic forum management 101. Of course trolls can be managed.
We have done this (before) by geoip country, region, useragent string, etc; because when you start to fuse data together (various easily measurable data on the user from standard web client information, like the $_SERVER superglobals, for example); you can score this and let the system auto moderate (keep the posts hidden) new users who scores are higher than the threshold you set based on your own “anti troll model”
The downside to this approach is that some people will be moderated who are legit and they will not have the “good feeling” of quickly posting and feeling welcome. This is basic detection theory 101.
However, these challenges can be overcome, to a high degree of confidence, by scoring the user based on a variety of metrics (geoip info, user agent string, IP address, initial pages visited, and other metrics).
Troll problems are easily solved if you follow an approach like this. You cannot have it both way. If you want to be “very open” then you will have trolls and trash. If you are too restrictive, you will have trouble building your community.
Everything is a trade-off and one size does not fit all. What works for Forum A may not be right for Forum B and so forth.
However, trolls can easily be controlled with a small about of “basic instrumentation” in place, scoring new users based your “anti troll algo” and then auto moderating (staging) those users until they are approved based on their behavior.
It’s basic detection theory…