The project is Quit within short training, when i train with a few multiple instances.

[Y0ungGun]

Describe the bug

When i start my RL Training, the below bug is logged within short training.

[INFO] Resuming from results\MyGrasp_250611\MyGrasp.
[INFO] Resuming training from step 3544.
[ERROR] UnityEnvironment worker 0: environment raised an unexpected exception.
Traceback (most recent call last):
File "C:\Users\smsla\anaconda3\envs\mlagents\lib\multiprocessing\connection.py", line 312, in _recv_bytes
nread, err = ov.GetOverlappedResult(True)
BrokenPipeError: [WinError 109] 파이프가 끝났습니다

To Reproduce
Steps to reproduce the behavior:

1. First, i open the socket server which inference my own CNN model, and get back the feature vector of image. And there is no any problem.

2. Secondly, i start the training whit command below, and start with unity editor.
   mlagents-learn config/250605.yaml --run-id=MyGrasp_250611 --time-scale=5 --resume

3. After few timesteps, i got the message above, and the project is quit like as i click the stop button in editor. In editor, there is no log or logerror.

4. If i start train with one or two instances, there is no error, a few instances such as 10 make error.

Console logs / stack traces
Please wrap in triple backticks (```) to make it easier to read.

Screenshots
If applicable, add screenshots to help explain your problem.

Environment (please complete the following information):

• Unity Version: [e.g. Unity 6000.0.30f1]
• OS + version: [e.g. Windows 11]
• ML-Agents version: ML-Agents 1.1.0(py), 3.0.0 in unity
• Torch version: 2.6.0+cu126
• Environment: (which example environment you used to reproduce the error)