Significant delay before first comment in VSCode (WSL) extension.

What version of the IDE extension are you using?

0.4.76

What subscription do you have?

Business

Which IDE are you using?

VSCode

What platform is your computer?

Linux 6.6.87.2-microsoft-standard-WSL2 x86_64 x86_64

What issue are you seeing?

I started experiencing severe performance degradation in Codex starting from version 5.2.

Regardless of the selected reasoning level (including extra-high), response times have become extremely long and inconsistent.

Example:

Previously:

• I asked Codex to generate a Python program (~2000 lines).
• It completed the task in ~25 minutes using extra-high reasoning.

Recently:

• I asked it to implement a small fix in the same project (now ~5000 lines).
• It generated ~100 lines of code.
• It took nearly 2 hours using maximum reasoning.

This behavior did not improve after upgrading to Codex 5.3.
The same issue occurs in both the IDE extension and the CLI.

There is no clear pattern to the slowdown.
It feels like requests are queued before processing.

In Codex 5.3, intermediate acknowledgements were introduced.
Normally, after sending a request, the extension outputs something like:
"Yes, I will review the code and find the issue."

However, in ~75% of cases, even this initial acknowledgement takes 3–10 minutes to appear, regardless of reasoning level.

During this time:

• The UI shows "Thinking..."
• No tool calls are visible
• No partial output is produced
• The extension appears idle

This happens even for small modifications.

What steps can reproduce the bug?

1. Open any project (even small or minimal)
2. Send a prompt with minimal context sometimes even without it (e.g. ask to explain a function or fix a small bug)
3. Observe delay before first acknowledgement (3–10 minutes)

Reasoning level does not affect the delay.

What is the expected behavior?

The extension should perform at least 5x faster.
Initial acknowledgement should appear within seconds, and overall response time should be significantly shorter and consistent.

Additional information

In some cases, the issue can be temporarily mitigated by re-sending the same request (i.e. steer it).However, this does not consistently resolve the problem, and the delay may reappear unpredictably.