It's important to realise the difference between consensus and Sybil prevention protocols. Consensus such as this one assumes participants are already selected and known to some extent. Then it runs its network messages to ensure it agrees on a value where everyone has the same value and won't revert or anything.

Sybil prevention is something that I wasn't taught at school and it concerns itself with creating the pre-requisites for consensus algorithm. Given the world population, how do we establish participant set for consensus while minimizing their chance to attack. Well, maybe by requiring them to waste record-breaking amounts of stupid compute.