[BUG] UI freezes during "thinking" phase; token count shows incorrect inflated values after recent update

Preflight Checklist

• [x] I have searched existing issues and this hasn't been reported yet
• [x] This is a single bug report (please file separate reports for different bugs)
• [x] I am using the latest version of Claude Code

What's Wrong?

Since a recent extension update, two related regressions have appeared during response generation:

1. UI freezes during thinking: The extension gets stuck in the "thinking" state with no streaming output. Previously, tokens would stream character by character in real time; now the UI hangs for several minutes.
2. Incorrect token count: The token counter displays inflated values (e.g. several thousand tokens) that do not reflect actual usage. The displayed count is significantly higher than what was consumed.

<img width="732" height="149" alt="Image" src="https://github.com/user-attachments/assets/a9626369-711c-4cb9-9f14-a54c7813706e" />

What Should Happen?

• Tokens should stream character by character during generation.
• The token counter should accurately reflect actual tokens consumed.

Error Messages/Logs

Steps to Reproduce

1. Open a conversation in the Claude Code sidebar.
2. Send any prompt that triggers extended thinking.
3. Observe that the UI freezes at the "thinking" indicator with no streaming text.
4. During the period of UI freezing, check the token count shown in the UI.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

Vscode Plugin: 2.1.195

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

VS Code integrated terminal

Additional Information

_No response_