I'm not sure exactly were they draw the line. Letters to the editor are up to the editor to choose, I'm sure of that. I'm not sure about comment sections in a news website, but they do censor those so I guess it's still in their domain, but I cannot be sure because someone would have to sue and (un)fortunately we don't sue as much as Americans. I did a duckduckgo, google and jusbrasil (a legal cases database) search and couldn't find anything relevant, it might exist, but must be buried very deep. It's all Facebook cases.
Foreign companies working in Brazil are bound by the same laws as everyone else here, so yes they forfeit the rights they have at home (like the right not to serve without explanation). The fact is that Facebook, Google, Twitter, etc (the companies that actually would get any attention), all have subsidiaries here that deal with their local legal problems. In theory any company serving a Brazilian is supposedly under our laws, but if the company does not have any subsidiary here, it would be very complicated to sue and hard to enforce any decisions.
The kind of thing that Twitter is removing is illegal here (terrorist incitement), so we won't see a Brazilian suing to get their account back. Plus, they can refuse service if they provide a valid explanation, like "he was using the platform for committing crimes".
I think the fact that companies cannot refuse service without explanation is what makes this approach "doable". We don't have to regulate the company as an utility or anything else, they are providing a communication service for people and if the users are not committing crimes with it they simply cannot refuse the service. If you publish an article on the Internet, you are not providing any services, there is no contract, so no protection for people's comments under it (I guess/assume).
I actually worked in a case were Google was being sued because of a blog on blogspot, their defense was on the lines of the "only the judiciary can decide what shouldn't be written" and "we are a service provider". They won, it was fast and cheap, a "no-brainer" type of case. I estimate they have spent from 15 to 40 USD in the case (I'm an office clerk in the judiciary, it's a good estimation).
I completely agree that the implementation is rather fraught with legal complexity (specially when it comes to comments on forums or foreign companies), there are plenty of things here that are like that. And I'm sorry I'm not sure of were they draw the line. Even being in the judiciary, that's not my line of work...
OTOH, consider that the American way were Twitter, Facebook and Google gets sued because of terrorists is also rather problematic.
While each company does have moderators that review content, The Next Web notes that it's a statistical impossibility to maintain that any company of such a size can review, or even find, all instances of offensive content.
So these companies are answering legally for things other people said. With our method you are only sued for what you say. And don't have to hire an army of censors just to make it look like you shouldn't be liable when things go wrong in life. If someone is offended by some content, get the urls and ask the judiciary, they are legally required to know what shouldn't be allowed. It's free to do that BTW, people don't need lawyers for simple cases like these.