nx_retry_policy.nx

nx_retry_policy.nx -- exponential backoff + jitter + max-attempts. module: nishi-core.ingest.retry_policy depends: nishi-core.io.syscalls, nishi-core.io.iso8601 disk_kb: 5 capability: CORE_IO license_tier: PUBLIC_NISHI_SUBSTRATE genealogy_id: aws_architecture_blog_exponential_backoff_jitter_2015 + polly_microsoft_resilience_library + google_sre_book_retry_amplification + nishi_ingestion_s_class_cardinal_2026 Retry primitive with exponential backoff + jitter + max-attempts. Composes against nx_circuit_breaker (trips when retries exhausted). Substrate Cardinal 14 graceful-degradation: every NX-INGEST adapter invokes this primitive to decide whether to retry a failed upstream call + how long to wait. ===== Why exponential backoff + jitter =========================== AWS Architectural Blog 2015: when many clients retry simultaneously (thundering herd), they amplify the upstream load just as upstream is recovering. Exponential backoff alone DOESN'T fix this — all clients still retry at the same exponentially-spaced times. JITTER (random offset within the backoff window) breaks the synchronization. Google SRE book: retry amplification is the #1 cause of cascading outages. Bounded max-attempts + jittered backoff prevents substrate from contributing to upstream failure. ===== Backoff formula ============================================ attempt_n: wait base * 2^n milliseconds, capped at max_wait_seconds jitter: rs_jitter_ms bounds the delay to [d/2, d] total: sum across attempts capped at attempts_cap ===== CORRECTED 2026-08-21 -- THIS BLOCK DESCRIBED A THING THE CODE DID NOT DO ====