[BUG] VS Code extension hides assistant messages after mid-turn API error + harness retry
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?
When viewing a past session in the VS Code extension transcript, an entire assistant turn (including its final text response and many tool calls) is invisible in the UI, even though the underlying .jsonl session file contains all the data correctly.
- Claude Code VS Code extension
- Linux (WSL2), Ubuntu
What Should Happen?
All assistant content from the same logical turn should render, including content produced after a transient API error and harness-driven retry.
Error Messages/Logs
Steps to Reproduce
The affected turn has this shape in the JSONL:
- Assistant issues several tool calls (e.g.
WebSearch). - Tool results return.
- An
API Error: The socket connection was closed unexpectedlymessage appears as a synthetic assistant entry. - The harness automatically retries; many more tool calls (
Write,Bash, etc.) succeed. - The assistant produces a final long-form text response.
In the VS Code transcript renderer, only items from step 1 are shown. Steps 2–5 are silently dropped. The user sees orphan tool-call entries with no follow-up, which looks like the assistant never replied.
Claude Model
Opus
Is this a regression?
I don't know
Last Working Version
_No response_
Claude Code Version
2.1.133
Platform
Anthropic API
Operating System
Ubuntu/Debian Linux
Terminal/Shell
WSL (Windows Subsystem for Linux)
Additional Information
Actual
Renderer appears to treat the synthetic API Error assistant entry as the end of the turn and stops drawing subsequent records.
This issue has 1 comment on GitHub. Read the full discussion on GitHub ↗