As we know mining has unfortunately become strongly centralised in Bitcoin. This has very damaging effects on the things that make Bitcoin valuable, both in a monetary sense and as a project trying to further the goals of crypto-anarchy.
Some allege some form of emergent consensus will help solve this, but I'm skeptical. The best-known proposal (BU) would almost certainly do the opposite, and it's not clear if there are viable alternatives that would work.
Consider the case of Ripple, specifically the new version of Ripple created by OpenCoin -> Ripple Labs -> Ripple. In the new Ripple validation is currently also very centralised, as there is a central core controlled by Ripple Inc.
(In Ryan Fugger's old Ripple things were even more centralised, as you had to conduct your tx on a ledger controlled by a single part, although you could choose which ledger and server to use).
In principle Ripple validation does not have to be centralised at all, as every validating node gets to choose its peers in a sort of trust relationship that is distinct from the trust graph related to the IOU system of rippling that is the heart of both the old and the new Ripple.
However, if people choose different set of validators there is a very real risk that they will not reach a global consensus so the ledger would fracture. Therefore basically everybody currently relies on a small set of servers run by Ripple Inc and its affiliates and this small set does not reciprocally trust other servers outside this set. In a way the emergent consensus has been to trust Ripple Inc.
In the long run this could change, because if there is a dense global mesh of mutually trusting validators with significant overlap then a global, distributed consensus would likely emerge. The big unanswered question is how to get there from here.
Similarly, a hub-and-spoke system for IOU trust relationships could draw in large numbers of people, who could then form P2P trust relationships, thus creating the originally intended mesh network of individuals who no longer have to rely on professional intermediaries. This appears still to be Ryan Fugger's hope.
TLDR: emergent consensus is a nice ideal, but there is a real risk of massive fracturing of the ledger and Ripple demonstrates that the fear of that may lead to total centralisation instead.