In a sybil attack, an adversary creates many fake identities/nodes and have them join the system. Computational puzzles have long been investigated as a possible sybil defense: nodes that fail to solve the puzzle in time will no longer be accepted by other nodes. However, a malicious node can behave in such a way that it is accepted by some honest nodes but not other honest nodes. This results in different honest nodes having different views on which set of nodes constitute the system. Such view divergence, unfortunately, breaks the overarching assumption required by many existing security protocols. Partly spurred by the growing popularity of Bitcoin, researchers have recently formalized the above view divergence problem and proposed interesting solutions (which we call view reconciliation protocols). All existing view reconciliation protocols so far have a similar Θ(N) time complexity, with N being the number of honest nodes in the system. As this paper's main contribution, we propose a novel view reconciliation protocol whose time complexity is only Θ(ln N/ln ln N). To achieve such an exponential improvement, we aggressively exploit randomization. The hidden constant factor in the asymptotic complexity of our protocol, however, is considerably larger than in previous protocols. Concrete numerical comparisons show that our protocol is more suitable for large-scale systems, while existing protocols are better for smaller-scale systems.