For simple trolls (as opposed to political actors), the dynamic is simple to model; you spend time and some effort to create accounts and say antagonistic things while people respond with abuse (to which the troll feels immune), and then harvest (via screenshots) examples of the saltiest tears for sharing with ones troll peers for lulz in other forums.

People do the exact same thing in real life, but it's more time-consuming and expensive to establish and maintain physical groups, both in economic terms and direct costs (legal or physical sanctions).

I do think you have a point, but also remember how public hate can be: slavery, segregation, genocide, Inquisition, witch trials, terrorism, violent crime, etc.

Unfortunately, humans have a history of willingly and publicly spreading hate without hiding behind text.

This is a huge topic with many possible solutions, but one option is to choose how we respond to it and how we let it affect us. Just knowing that someone might be more aggressive through text than in person might help to dampen the impact we let text have on us.