FASE 37B-GRAPH-RUNTIME-CORRELATION-01 cruza señales de topología (GitNexus graph reasoning) con señales runtime observacionales:

• Graph hotspots / blast radius / governance risk (GitNexus topology)
• Cognitive Health Layer (37A): health_score, routing_confidence, nodes
• SLO: overall_status, violations_total
• Autonomous triage: severidad agregada
• Federation guards: estado y counters
• Evidence lineage: replay/stale/invalid

El resultado permite afirmar (con disciplina epistemológica):

“Este hotspot arquitectónico coincide con una degradación runtime real”.

• No cambia routing.
• No ejecuta remediación.
• No escribe en runtime/state/*.
• No hace scans infinitos ni polling agresivo.
• No convierte inferencias en hard facts.

• hard_facts: lecturas directas de snapshots/metrics disponibles.
• inferred: correlaciones derivadas (p.ej. fan-in alto + replay detections + health warning).
• unknowns: fuentes ausentes o no accesibles.

La fase expone explícitamente unknowns y unavailable_fields.

Score determinista 0.0–1.0 basado en weighted sum de:

• centrality/fan-in/fan-out
• blast radius + governance risk
• health degradation + routing confidence degradation
• SLO status
• triage severity
• federation guard state
• evidence replay/stale/invalid

Clasificación:

• 0.00–0.24 INFO
• 0.25–0.49 LOW
• 0.50–0.69 MEDIUM
• 0.70–0.84 HIGH
• 0.85–1.00 CRITICAL

• GET /runtime/correlation
• GET /runtime/correlation/summary
• GET /runtime/correlation/hotspots
• GET /runtime/correlation/blast-radius
• GET /runtime/correlation/findings
• GET /runtime/correlation/recommendations

Expuestas en GET /metrics (gateway):

• ailab_correlation_score
• ailab_correlation_hotspots_total
• ailab_correlation_high_risk_total
• ailab_correlation_critical_total
• ailab_correlation_unknowns_total
• ailab_correlation_recommendations_total
• ailab_correlation_runtime_health_linked_total
• ailab_correlation_graph_health_linked_total