API stream timeout causes silent failure - no retry, no user notification

Bug Description

When the Anthropic API times out mid-stream during a Claude Code session, the tool silently goes idle with no indication to the user that anything went wrong. There is no retry attempt and no error message displayed.

Expected Behavior

When an API stream times out, Claude Code should either:

1. Automatically retry the request
2. Or at minimum, display an error message to the user indicating the failure

Actual Behavior

Claude Code silently enters an idle state. The user sees nothing - no error, no prompt, no indication that the previous operation failed. The session just stops responding.

Error Message

API Error: Stream idle timeout - partial response received

This error appears in internal logs but is not surfaced to the user.

Reproduction

This has occurred multiple times in our environment:

• During long-running agent operations
• When spawning multiple background subagents
• The API connection stalls, eventually times out, and Claude Code goes silent

Impact

In our case, a host session went silent for 8 hours after hitting this error. The user had no way to know something went wrong until they manually checked in.

Environment

• Claude Code CLI
• Running in tmux sessions
• Long-running agent workflows with subagent dispatches