Chain co-failure graph

28 of 118 reproduced techniques recur across chains

The honest default. 20 chain pairs carry more than one shared technique. Every edge shown is backed by at least two independent techniques, so it is not one reproduction campaign drawing a shape.

Read this one with suspicion. 149 pairs, but 129 carry exactly one technique. A technique reproduced on N chains emits N*(N-1)/2 pairs at once, so density here can mean "one technique touched a lot of chains" rather than "these chains are related".

The version that infers nothing. One line per real reproduction: 28 techniques against 36 chains. Less shapely than the projection and strictly more truthful.

A chain is a deployment, not an implementation. These edges are an upper bound on independence, never a measurement of it - two chains can share an edge because they vendor the same library. Lineage is curated for one technique so far, so the registry reports what it can measure and declines to imply the rest.