Authenticated callers can provide a request chat template when no server-side `--chat-template` is configured. Additionally, a malicious model repository may provide a template that is compiled and rendered by the `apply chat template()` function without a wall-clock timeout, output-size limit, or iteration bound. Because rendering occurs in a `ThreadPoolExecutor` with one worker by default, a nonterminating render can block the `/v1/responses` endpoint and other users. Furthermore, an output-expanding render can lead to server memory exhaustion.