My reading of Skycoin whitepapers Sky: an Opinion Dynamics Framework and Model for Consensus over P2P Network (WP1) and A Distributed Consensus Algorithm for Cryptocurrency Networks (WP2) certainly did not leave me an impression of its consensus system being a gem.

WP2's assumption (0.3.2 page 2) that node is somehow more than a procedural engine capable of generating fresh content.

Node is a content generator. The node is able to receive raw data (e.g. low-level, elementary events such as transactions) and produce an independent research that leads to a new opinion (e.g. block hash).

The entire analysis (in WP1) is done using mean field theory perhaps under the assumption that the mean behavior of the nodes fully defines the desired (and undesired) properties of the system. Unfortunately, it is the edge cases, not the mean behavior, that are the most critical to a system like this.

And even without using the analysis for all the edge cases, all the protocol can offer is a mere 13% failure tolerance.

Theoretic analysis and simulations both show that it can tolerant failures by at least 13% random nodes ...

Note that it is 13% of nodes not stake. Also note that the nodes are rather inexpensive to create (at a cost of approximately $70/hour per 10,000 nodes).

I could be wrong but I am completely missing the point or usefulness of this protocol.