I admit this may sound a bit odd (and possibly even mildy schizophrenic), but I adopted the shell-kernel model for interactions. Now, when someone says something I agree or disagree with, it goes first through the shell, giving me some time to filter it, analyze it, and consider whether or not a response is really required. If the shell says, hey, take a look at this, only then does the kernel respond. The result is that most of the time the kernel stays calm, cool, relaxed and can focus on important things (and also, ensure any responses are polite rather than overly aggressive or confrontational).
It was a bit tricky at first, because people might think I'm ignoring them (or even somewhat autistic), so the shell has to be personable and pay attention and act like a normal human being. The real benefit, however, is that the whole 'triggered' thing (rather prevalent in today's society) just stops being a problem, and you get a little breathing space in which to think about how best to respond to someone.