Nick Malik in his blog "Correct" is a point of view makes the critical point that when we state that enterprise architecture is about "doing the right thing" we must not gold plate the process, we must aim to do a good job rather than a perfect job.
Enterprise architects often complain about the business direction being unclear. As I stated in my blog, The Absence of Strategy, this should never be an excuse. You can always derive sufficient context for any architectural decision to to be good enough.
In hindsight, many decisions seem wrong and have created a legacy. But at the time they may well have been right. The alternative of delay or do nothing may have been much worse. Failing to meet regulation, delaying product release or missing peak trading could have much greater costs than unpicking the legacy of hasty actions.
Enterprise architecture does not need to be what Nick Malik calls the "ultimate BDUF exercise". There are a range of enterprise architecture management styles that suit different organizational contexts
Sometimes these decisions could have been made better. Why weren't they? Absence of information is usually the answer. We need to be out there listening, learning and informing for our point of view to be heard. Decision processes exist, we need to be part of them rather than create new ones, we need to be part of the context of business decision making.