Sources & References — AG08:2026

The primary specification for this vulnerability. Defines cascading failure vulnerability in multi-agent systems, describes propagation mechanisms, and lists prevention strategies including circuit breakers, graceful degradation, output validation, and blast radius limitation. All module content is based on this specification.

The foundational reference for production resilience patterns in distributed systems. Chapter 5 (Stability Patterns) documents the circuit breaker, bulkhead, timeout, and fail fast patterns described in MIT02 and MIT05. Nygard's formalization of the circuit breaker pattern — including the three states (closed, open, half-open) — is the standard reference cited by all subsequent implementations.

Martin Fowler's widely-cited explanation of the circuit breaker pattern, documenting the state machine (closed → open → half-open) that governs circuit breaker behavior. The three-state model described in MIT02 (slide 18) is drawn from this reference.

Netflix's documentation of their cascading failure experiences and the Hystrix circuit breaker library developed to address them. The thundering herd and retry storm patterns described in Pattern 4 and Scenario 3 are directly derived from Netflix's documented production failure modes.

The definitive explanation of why jitter (randomized retry timing) is necessary in addition to pure exponential backoff. Documents how synchronized retry bursts from multiple clients repeatedly overwhelm recovering services — the thundering herd pattern shown in Scenario 3. MIT04 is drawn directly from this reference.

Academic documentation of the chaos engineering discipline pioneered at Netflix, including the Chaos Monkey tool. Provides the theoretical and practical basis for MIT09 (chaos engineering and failure mode testing). The paper documents the principle that resilience mechanisms not tested under realistic conditions should not be trusted in production.

Comprehensive reference for chaos engineering practice, including how to design failure injection experiments and what to measure. The failure scenarios in this module (Scenarios 3 and 5) are the type of failure modes that chaos engineering is designed to discover before they occur in production.

Google's Site Reliability Engineering documentation of cascading failure patterns at scale, including load shedding, graceful degradation, and feedback loop prevention strategies. The three-zone defense model (Zone 1: detection, Zone 2: containment, Zone 3: blast radius limitation) in slide 22 is informed by this material.

Covers fault isolation, health monitoring, and resilience architecture for distributed service-based systems. The bulkhead isolation pattern (MIT05) and monitoring independence requirement (MIT08 Takeaway 6) are consistent with the defense-in-depth architecture described in this publication.

Covers workload resilience patterns including fault isolation, retry policies, graceful degradation, and failure mode testing. The SQS queue isolation patterns relevant to Scenario 3 (retry storm) and the idempotency token pattern (MIT07) are described in this reference.