Glossary (defined once; questions reference by name)

• At-least-once delivery — messages may be delivered more than once, never lost; consumers must dedupe. (distributed-systems-01, messaging-events-01)
• Atomic phase — a step of a multi-step request done in one DB transaction with a recovery point, so a retry resumes safely. (distributed-systems-01)
• Architecture decision record (ADR) — short doc in the repo: context, decision, consequences of one significant choice. (leadership-decisions-01)
• Backoff with jitter — retry with exponentially growing, randomized delays to avoid synchronized retry storms. (system-design-01, distributed-systems-01)
• Bounded context — DDD boundary within which a domain model is consistent; natural service seam. (leadership-decisions-01)
• Burn rate — how fast the error budget is consumed relative to the SLO window; 1x = exactly exhausts it at window end. (reliability-observability-01)
• Change data capture (CDC) — streaming row changes from a database's log (e.g. Debezium on the WAL) to other systems. (messaging-events-01)
• Distributed monolith — services that must deploy together or share a DB; costs of both, benefits of neither. (leadership-decisions-01)
• Dual write — writing to two systems (DB + broker) without a shared transaction; one can succeed while the other fails. (messaging-events-01)
• Error budget — 1 − SLO; the allowed unreliability that can be "spent" on change. (reliability-observability-01)
• Error budget policy — agreed actions (e.g. launch freeze) when the budget is exhausted. (reliability-observability-01)
• Fail open — when a protective component fails, allow traffic rather than block it. (system-design-01)
• Idempotency key — client-generated unique id per operation so retries are recognized and not re-executed. (distributed-systems-01)
• Load shedding — deliberately dropping lower-priority work under overload to keep the system up. (system-design-01)
• Modular monolith — one deployable with strongly enforced internal module boundaries. (leadership-decisions-01)
• Saga — sequence of local transactions with compensating actions instead of a distributed transaction. (messaging-events-01)
• SLI — service level indicator: measured ratio of good events to valid events. (reliability-observability-01)
• SLO — target value for an SLI over a window, e.g. 99.9% over 28 days. (reliability-observability-01)
• Sliding window counter — rate-limit approximation weighting the previous fixed window's count by overlap. (system-design-01)
• Strangler fig pattern — incrementally replace a legacy system behind a facade until it can be retired. (leadership-decisions-01)
• Token bucket — tokens refill at a fixed rate up to a cap; each request spends one; permits bursts. (system-design-01)
• Transactional outbox — write events to an outbox table in the same transaction as the business data, relay later. (messaging-events-01)