The DAO did exactly, exactly, what it was supposed to do. In no way can it be have said to fail, except in the "wasn't a great investment" sense. Maybe in failed in the "wow using terrible languages to write complicated code is a bad idea" way, too.
Ethereum failed, as they decided smart contracts aren't a great idea, so they're now going to involve humans when they feel like it. The whole premise of Ethereum was the opposite of what they're doing.
I believe that Ethereum+DAO was a sucessful experiment (with negative results) - they had a vision/hypothesis that smart contracts without human/legal overrides are a good idea, despite the experience of the whole financial and legal world over the last couple of millenia telling them that this vision is naively utopic. The whole point of Ethereum was implementing and testing this vision in practice, which they did.
In practice they found out (and showed the public) the kind of consequences everyone was expecting; and so we have a successful experiment giving some evidence on whether smart contracts w/o overrides are a good idea - apparently, it's not; and now based on this we can build better systems that properly account for the risks that DAO had.
Ethereum failed, as they decided smart contracts aren't a great idea, so they're now going to involve humans when they feel like it. The whole premise of Ethereum was the opposite of what they're doing.