Operational Memory Architecture for Kubernetes: Preserving Causal Context Across the Evidence Horizon

The Operational Memory Architecture (OMA) is an open-source framework for preserving causal diagnostic context in Kubernetes clusters before it is destroyed by the platform's native event retention mechanisms. OMA introduces the evidence horizon — a boundary (~90 seconds) after which LastTerminationState data is permanently overwritten — and addresses it through three causal patterns (P001: OOMKill chain, P002: ConfigMap env var misconfiguration, P003: ConfigMap volume mount propagation) implemented as a Go-based collector, SQLite operational memory store, and canonical query interface. Validated on Minikube and AKS 1.32.10 with 30-run statistical analysis and concurrent stress evaluation. Full implementation at https://github.com/opscart/k8s-causal-memory