Measurement infrastructure at scale: distributed-systems lessons

One thing I find under-discussed in the merit-measurement conversation is how much it shares with distributed systems engineering. You have multiple sources of truth, each partially correct, each with latency and noise, and you need to reconcile them into a coherent state without losing fidelity. The Risk Efficacy pillars map surprisingly well to consensus algorithms — calibration is essentially a prediction-confidence vote, and the integration step is a quorum read. Has anyone written this up properly?