Explore the connection between transport selection, recoverable event processing, and bounded retries. These guides treat gaps, repeated observations, timeouts, and stale caches as normal states to design, not as surprises to hide. Start with the transport comparison, then follow the path toward a replayable index or a smaller request policy. The right amount of machinery depends on your workload and the guarantees your interface actually promises. Follow the articles below in publication order or choose the guide closest to your current feature. For related definitions, keep the API glossary nearby; for source-specific details, use the primary reference library. Every article includes its own next steps, so this topic page can work as a starting point or as a way to revisit a particular implementation decision.
Related articles
JSON-RPC, REST, or WebSockets?
Match your request pattern to the feature, then design reconnection and recovery before scaling.
Read the guide
Build an event index that can recover
Design checkpoints, replay, reorganization handling, and historical coverage as first-class features.
Read the guide
Rate limits and retries, minus the chaos
Classify failures, respect rate-limit signals, and put every operation on a bounded recovery budget.
Read the guide