Network Health & Integrity
Selection-bias-aware QoS quality scoring across all indexers the gateway routed traffic to. Ranks on actual service quality — Wilson-reliability, latency, freshness — not raw query volume, and flags indexers holding allocations the gateway routes around.
Behaviorally Correlated Clusters
Correlation is not common control. These groups share behavioural signals (allocation overlap + registration cohort or identical cut parameters) — a working hypothesis for human review, not an accusation. Shared infrastructure providers produce correlated signatures legitimately. No punitive action is taken on this basis. Tier 3 (shared on-chain funding) is intentionally not shown until funding-graph analysis ships.
No behaviorally-correlated clusters detected at current thresholds.
Q-Score = Wilson-reliability × latency-decay × freshness (weighted product), EWMA-decayed over 30 days, normalised per-deployment, weighted by served-query share. Gap= allocation share minus served-query share (high positive = the gateway routes around this indexer despite its allocations). Source: QoS Oracle V1. Informational; absence of routed data ≠ absence of problems. Clustering & crowding-out analysis coming next.