But anyway, I went ahead and checked out the white paper of nexus mutual [1], because I was curious. It appears to have a serious amount of hand-waving on one of the most important topics: the risk correlation between offered insurance products. They do reference the correlation matrix, but the only mention of how the value of the matrix is determined is to say that if independence between cells can be assumed, the math is very simple. Doing a quick search through their website and code, it looks like they are indeed just assuming that products aren't correlated, instead of trying to estimate real correlation values. This means that their minimum capital requirements (and thus the implied risk of default) are incorrectly calculated if that assumption is violated--which it certainly is.
This seems to be by design, and baked into the incentive structure of the whole concept. There's just no practical way to crowdsource proper correlation evaluation and adjustment, and the economics stop being remotely competitive if you just guess at correlations and treat it as another risk to be hedged. You can see this later on in the white paper (appendix A), where they point out that the economic viability of the project depends on lowered labour costs because product creation, assessment, and policy issuance are crowdsourced or automated. Their game theory/tokenomics are focused on providing incentives for individual products to price risk accurately, however: they do not propose any mechanism for adjusting or calculating MCR based on correlation risk when new products/coverage are offered. This is further evidence that they're just assuming independence without any real justification, and thus being chronically undercapitalized.
[1] https://nexusmutual.io/assets/docs/nmx_white_paperv2_3.pdf