Rate Limits and Failover

Endpoint failures are normal production events. Design for them upfront.

Rate-Limit Planning

assume public infrastructure has limits
avoid bursty polling patterns when indexed alternatives exist
separate heavy history workloads from simple chain reads where possible

Failover Planning

keep more than one endpoint available for critical workloads
decide which calls can be retried automatically
surface degraded service clearly in your app
log endpoint-specific failures so you can spot provider issues quickly

Avoid These Assumptions

that all providers expose the same features
that every public endpoint is suitable for production traffic
that write and read traffic should always use the same backend