Back to Docs
Developer Docs

Retry, Backoff, and Queueing

Production resilience strategy for matching calls using retry policies, exponential backoff, and queueing.

Reliable automation requires controlled retries and queueing. This page provides a practical policy for matching workflows that may hit transient limits.

Retry decision matrix

  • Retry: 429, 5xx
  • Do not retry blindly: 400, 401, 403, 422

Backoff pattern

attempt = 0
delay = 500ms
while attempt < maxAttempts:
  call API
  if success: break
  if retriable(status): sleep(delay + jitter)
  else: fail fast
  delay *= 2
  attempt += 1

Queueing strategy

  • Prioritize urgent client-facing requests
  • Queue low-priority matching jobs near quota edges
  • Use dead-letter queue for repeated permanent failures

Implementation checklist

  • Add retry cap to avoid infinite loops
  • Add jitter to reduce synchronized spikes
  • Tag queued jobs with priority and expiry
  • Record terminal failure reason for support and QA