gw-06 — References

Foundational reading

Google SRE Book — "Handling Overload," "Addressing Cascading Failures," "Load Balancing in the Datacenter." The canonical treatment of retries, budgets, shedding, and graceful degradation. https://sre.google/sre-book/handling-overload/
Marc Brooker — Timeouts, retries, and backoff with jitter (AWS Builders' Library) and the metastable-failure posts. The retry-storm and jitter canon. https://aws.amazon.com/builders-library/timeouts-retries-and-backoff-with-jitter/
Metastable Failures in Distributed Systems (HotOS 2021, Bronson et al.) — the formal framing of self-sustaining overload.

Mitzenmacher, The Power of Two Choices in Randomized Load Balancing — the P2C result (exponential reduction in max load).
Twitter Finagle — P2C + EWMA ("p2c / leastLoaded") and the Aperture subsetting balancer. https://twitter.github.io/finagle/guide/Clients.html
Envoy LB docs — P2C (LEAST_REQUEST), outlier detection, zone-aware routing, circuit breakers. https://www.envoyproxy.io/docs/envoy/latest/intro/arch_overview/upstream/load_balancing/load_balancers

Netflix concurrency-limits (Java) — the Gradient/Vegas-style adaptive limiter; read the README and the Limit implementations. https://github.com/Netflix/concurrency-limits
Netflix TechBlog — Performance Under Load (adaptive concurrency limits at the edge). https://netflixtechblog.medium.com/performance-under-load-3e6fa9a60581
Envoy adaptive concurrency filter (the same idea in the data plane).
TCP Vegas / congestion-control papers — the control-loop intuition.

Michael Nygard, Release It! — circuit breakers, bulkheads, the stability patterns vocabulary.
Netflix Hystrix (archived) + resilience4j — circuit breaker / bulkhead / rate limiter implementations to read.

Upstream: gw-03 (endpoint phase + timeout), gw-04 (subset to balance over).
Downstream: gw-08 (Envoy implements all of this via xDS), gw-11 (the metrics that make it observable), gw-05 (reconnect backoff).