Hi @Lorimer I think your observation is correct that there should be enough nodes to Nakamoto coefficient of 3 per country (since country limit is 2 for most subnets, which would mean 5 nodes need to collude to take over a 13 node subnet, which would be 2 in same country + another 2 in same country + any other node, which means Nakamoto coefficient is 3). Otherwise the IC target topology would not have been reached.
But to support the statement of @sat, country is the most limiting factor in decentralization (there is much more variety in Node Providers, cities, data centers, and data center operators than countries), and spare nodes are not taken into account in the optimization tool, so suboptimization by seeking the maximum decentralization for one subnet needs to be avoided.
In the example above, swapping two nodes (for example adding one node from a Node Provider in South Africa, next to replacing the unhealthy node) is indeed interesting but does not change the nakamoto coefficient, as Sasa also indicated. With 3 (instead of 4) US nodes, and 2 nodes in Switzerland you still only need to have control over these two countries in order to take over the subnet. What is more interesting is the add another - third - node that replaces one American or Swiss node, then the country limit would bump of from 2 to 3. The current dre-tool can be run multiple times so that you can continuously improve each subnet, but for the two reasons above (country limit, and need for enough spare nodes) it’s better to only do this increased optimization for the larger (i.e. high priority) subnets such as NNS, SNS, II and Fiduciary subnets.
Please have a look at the dre-tooling and happy to hear your throughts.