For example, Buffer could have easily added a policy saying that each employee's compensation may be adjusted by, say, +/- 20% based on individual factors. That would give some uncertainty, and thus a bit of individual privacy. Peers would still have some confidence that they were, more or less, in a similar range as others with their public performance characteristics.
Let's put ideology aside (i.e. do we want this to work, according to our theories of human nature) and focus on the actual effects. How do you know if you've succeeded? How do you measure this? How do you design the experiment?
Call me skeptical, but I can't help but think that Buffer is doing this, largely, to say "look at us!" and "this makes an interesting blog post!".