Among other things, we captured a 100% set of tweets surrounding the election in Iran in 09, and released a paper on them. It was interesting, because it ended up all being a bit of a precursor to the Arab Spring and many current events in the middle east today.
I loved doing it. The API team was responsive, and although we were never able to get full Firehose (despite offerring large amount of money) we were able to easily play with and access a lot of data.
Much/all of that has changed. The API has gotten really complex to work with. The terms of service are no longer friendly. Getting whitelisted accounts for standard API calls is impossible, and overall feeling like we have any influence or use to Twitter is dwindling.
Whereas at one point I felt it was a joy to work with, now I feel its impossible. Getting the data to study is impossible (or expensive, since a few companies control the firehose), and there's the constant feeling of getting shut down at any moment. Plus, there's just too much data to grab given the clunky methods you have to use to do it. I'd kill for a hard drive of the 'network graph' with diffs over time. As-is, its impossible to answer certain academic questions.
To a certain degree, while Twitter is used far more now than in 2009, it feels boring, obvious and less important to me. So at the end of the day, nothing of value was lost, but I do miss it as it was interesting at one point.