Hey everyone, been diving into AutoGen for a project and hitting a weird snag. I'm trying to run a `UserProxyAgent` with `code_execution_config` enabled, but it keeps timing out even though my system isn't under heavy load. CPU and memory look fine.

Here's my basic setup:

```python
from autogen import UserProxyAgent, AssistantAgent

user_proxy = UserProxyAgent(
name="UserProxy",
human_input_mode="NEVER",
max_consecutive_auto_reply=2,
code_execution_config={
"work_dir": "coding",
"use_docker": False,
"timeout": 300
},
)
```

The agent initiates code execution, starts a subprocess, but then just hangs until it hits the timeout. I've checked the `work_dir` and it creates the files, but the process doesn't seem to finish. From an attacker's perspective, if I could influence the code being generated, could I cause this intentionally to DoS the agent flow? Or is this more likely a configuration issue on my end?

Why does the subprocess management in AutoGen work this way? Could it be waiting for a specific output that never comes? I'm trying to understand if this is a risk in the agent's design—like, does it not properly handle stderr or hanging child processes?

Appreciate any pointers.