Keiko was functionally free beginning in the summer of 1999, allowed to roam at will out of his seapen, but returning voluntarily until that day in August 2002 when he hooked up with a pod of wild orcas and never came back, showing up in Norway instead and reestablishing contact with humans. The Keiko experiment was not a failure except in reaching a final goal that the industry had a direct hand in ensuring was never reached -- namely, a positive identification of his familial pod so he could be reunited with them. What we learned from Keiko is that such identification is vital to a complete reintegration.
But in every other regard, this was a successful reintroduction to the wild. He learned to feed himself. He was independent. He clearly appeared to be healthy and happy, right up until just before he died. And the lung infection he died from may well have been contracted in captivity anyway.
I quote the famed whale researcher Paul Spong on this subject in my forthcoming book, Of Orcas and Men: What Killer Whales Can Teach Us. Here's what he says: