For browser automation, scraping, or API calls through a proxy, timeout and retry settings strongly affect stability. Too much concurrency, too-short timeouts, or endless retries can slow jobs and create many failures.
Baseline settings
- Connect timeout: start around 10–30 seconds.
- Read timeout: match target site speed; slow sites may need more time.
- Retries: usually 2–3; avoid unlimited retries.
- Concurrency: start small, then raise based on success rate.
When to retry
Transient network blips, brief target outages, or a single egress failure can use limited retries. Add a short wait between attempts so you do not hammer the target or your own workers.
When not to retry
- Wrong username or password
- Expired plan or insufficient balance/traffic
- Wrong protocol or port
Fix the config instead of retrying — retries only amplify the error.
Debug flow
- Prove one proxy with cURL.
- Run a tiny job at 1–5 concurrency.
- Record success rate, latency, and failure reasons.
- If stable, raise concurrency gradually.
Stable proxy jobs usually find a good cadence for the target, protocol, and plan — not max concurrency on day one.