# Step 3 — Implementation Detail

1. Define RunResult with run ID, status, count, errors, and cursor.
2. Select full or incremental connector execution from the requested mode.
3. Publish successful documents before returning the new cursor.
4. Retry connector/publisher failures up to max_retries.
5. Return a dead-letter-compatible result after final failure.
6. Dispatch webhook payloads to the configured connector.
7. Add orchestration and fixed-run-ID tests.