It seems like this would be a great case for doing a machine learning approach. Something good at anomaly detection like single class support vector machines, perhaps.
I wonder how well that would translate for a client services person, though. "Why do we think this client is off?" "Because black box."
I would argue that you are correct, but I wouldn't limit it to Twitter. There are many crackhouses -- even hacker news can be one (though thankfully, our junkies remain civil and politely die in a corner where no one sees them). It is never as easy to be inhuman to someone as when hiding behind the veil of apparent anonymity.
I seem to recall, a while back the creator of RVM suffered a similarly disheartening setback when rbenv came out. RVM is still going strong. The community has not decided on lightweight frameworks. The choices you make won't be the same as lotus. You will no doubt attract a group of like-minded people who will be grateful for your contributions.
There isn't even much technical ingenuity here. I'm not sure how this is even noteworthy. You save $1 between two people? So you split it for a lousy $0.50 or maybe $100 per year? Just stop drinking lattes and you'll save that in a month.
Not to mention if you live further away, you deserve to pay more to make up for the extra resources needed to move you around, as nchuhoai said.