Technical Field Notes
Analyses of distributed systems realities: failure modes, edge state consensus, append-only synchronization, and production survivability.
04 // Document Index of Engineering Notes
3 DOCUMENTS| Ref | Title & Core Invariant | Topics & System Primitives | Date | Length | Access |
|---|---|---|---|---|---|
| 01 | At-least-once delivery is easy; exactly-once is a product decision Why idempotency keys, explicit device-side ACK contracts, and retry budgets are the only reliable way to prevent duplicate execution when the physical network partitions. | Idempotency Keys Device ACK Retry Budgets Dead-Letter Queues BullMQ & SSE | OCT 2025 | 6 min read | Read → |
| 02 | Offline-first is a data-model problem, not a sync button A sync button is a UX admission of defeat. True resilience requires append-only local mutation queues, schema-per-tenant isolation, tombstones, and causal reconciliation. | Local Queues Schema-per-Tenant Causal Ordering Tombstones PostgreSQL | AUG 2025 | 8 min read | Read → |
| 03 | Designing for the six-hour outage What happens to order streams when an edge kitchen loses power for six hours? Building heartbeat telemetry, exponential backoff ceilings, and disaster-recovery snapshots. | Heartbeat Pipelines AWS IoT Core DynamoDB Snapshots Backpressure Runbooks | JUN 2025 | 7 min read | Read → |
These field notes are derived directly from verified production incidents, architectural postmortems, and load-testing verifications. None of these essays describe theoretical frameworks; they document how systems behave when networks partition, power drops, or duplicate transactions flood a service mesh.