Significant Performance Regression in Codex from v0.4.39(VSCode) to v0.4.40+(VSCode) on Specific Repository

What version of Codex is running?

codex-cli 0.58.0-alpha.10

What subscription do you have?

ChatGPT Plus

Which model were you using?

GPT-5 and GPT-5.1

What platform is your computer?

Linux 6.8.0-87-generic x86_64 x86_64

What issue are you seeing?

I encountered a significant performance regression when running the same task through Codex on Ubuntu using VSCode. The repository contains both C++ and Python code.

Here are the execution times I observed:

• Codex (VSCode) 0.4.39

binary version: codex-cli 0.58.0-alpha.1
Model: GPT-5
Execution time: ~2 minutes

• Codex (VSCode) 0.4.40

binary version: codex-cli 0.58.0-alpha.10
Model: GPT-5
Execution time: ~36 minutes

• Codex (VSCode) 0.4.43

binary version: codex-cli 0.58.0-alpha.10
Model: GPT-5.1
Execution time: ~42 minutes

Even though the task is exactly the same, the processing time increases drastically starting from version 0.4.40.

This issue does not occur on all repositories, but it consistently happens on a specific repository that includes both C++ and Python code.

What steps can reproduce the bug?

It run the same task through Codex on Ubuntu using VSCode with v0.4.39 and v0.4.40.
The repository contains both C++ and Python code.

What is the expected behavior?

Execution time should remain similar across versions when running the same task, or at least not degrade drastically.

Additional information

_No response_