The engine resolves Vertex AI redirect URLs in parallel using HEAD requests to extract the final destination from the Location header without loading full page contents.