Skip to main content

Current guidance

Non-admin API clients should assume conservative retry behavior:
  • Retry transient failures (5xx, transport timeouts) with exponential backoff.
  • Do not blind-retry writes unless route behavior is idempotent or guarded by server checks.
  • Avoid concurrent duplicate writes for the same user action.

Stream endpoints

For SSE endpoints (leaderboard and locker room streams):
  • reconnect on disconnect with short backoff,
  • resume by re-fetching latest snapshot endpoints as needed.

Operational best practice

Use client-side request deduplication and bounded retry budgets to prevent cascading failures.