The only workable concept of consensus I've ever discovered is stability. If hundreds of people edit a piece of work in good faith over a long period, what changes least over time may be presumed to be there by consensus. However even the most apparently stable elements of a work may be deposed quite easily. The result may be a new consensus or, in other cases, a period of instability where the new version and the old version compete.
As an editing community, I think we're probably far more tolerant of radical change than we used to be, but probably not tolerant enough. Most of us (myself included) are somewhat squeamish about junking bad stuff and starting over, and this squeamishness is probably to the detriment of the project.
On policy issues, the overriding requirements and purpose of the project are often ignored or poorly understood by large parts of the community, which I believe is mostly due to lack of acculturation in the core values. The antithesis of this is the statement that the encyclopedia takes precedence over the community. This encyclopedia-community dichotomy makes consensus-building difficult in areas of developing policy. However the community quickly falls into line with new policy once it can be shown to work. In that sense, policy is what works.