Reliability

Reliability determines whether your Engine consistently receives bounties and whether your results arrive in time to participate in rewards. A reliable Engine is predictable, responsive, and safe under load.

Reliability Criteria

Your Engine should:

• Accept webhooks consistently over HTTPS
• Validate signatures on every request
• Return 202 Accepted quickly
• Process bounties asynchronously in workers
• Post assertions back successfully and on time
• Degrade safely (UNKNOWN instead of crashing)

Common Failure Points

• Webhook endpoint is not reachable (DNS, TLS, firewall)
• Signature validation fails (secret mismatch)
• Web server blocks on scanning (no queue, no worker separation)
• Worker crashes on edge cases (empty bounty, unsupported type)
• Artifact downloads fail intermittently (network, timeouts)
• Concurrency is too low for configured rate limit
• External tools hang (no timeouts)

Minimum Operational Checklist

Availability

• Endpoint is publicly reachable over HTTPS
• Certificates are valid and renewed automatically
• Health check endpoint exists (optional but recommended)

Performance

• Webhook request returns quickly (do not scan in the request thread)
• Worker concurrency supports your configured rate limit

• Timeouts are enforced for:

  • artifact download
  • external tool execution
  • response submission

Correct behavior under stress

• Unsupported artifact types return UNKNOWN with bid 0
• Empty or malformed bounties do not crash the service
• Retries do not create duplicate assertions (idempotency where possible)

Logging and Observability

At minimum log:

• Bounty received (id, type)
• Signature validation result
• Job queued and started
• Analysis finished (verdict, duration)
• Assertion post result (success or error)

Protecting against “late results”

Late results often earn nothing. Reduce “late” outcomes by:

• Enforcing timeouts and returning UNKNOWN when needed
• Keeping artifact downloads fast (bandwidth, caching if appropriate)
• Avoiding slow startup paths and cold starts
• Not overloading workers beyond capacity

If your Engine is marked Failed

If your Engine enters a Failed state:

• Fix the underlying reliability issue first
• Re-test in the Development Community if needed
• Follow your verification recovery process as required by your workflow