claude --continue freezes on drive root (D:\) due to unhandled ripgrep timeout in resume path

Bug Description

claude --continue (or claude -c) freezes when run from a drive root (e.g., D:\). The UI renders but input is completely frozen — cannot type, Ctrl+C doesn't work. New sessions (claude) from the same directory work fine.

Steps to Reproduce

1. Open terminal, cd D:\ (any drive root with many files)
2. Run claude — works fine, have a conversation, exit
3. Run claude -c to resume — UI shows the conversation but input is frozen

Root Cause (from debug logs)

On startup, Claude Code scans the working directory for CLAUDE.md:

claude.exe --ripgrep --files --glob CLAUDE.md --sort=modified --no-ignore --hidden D:\

On a drive root with 190K+ files, this times out after 20 seconds with RipgrepTimeoutError.

Critical: The new session code path recovers from this error and continues normally. The resume code path does not — it goes completely silent after the error, and the input handler never initializes.

Debug log evidence

• New session: rg AbortError at 06:45:30 → session continues, assistant responds at 06:46:23 ✓
• Resume: rg AbortError at 06:46:47 → last log entry at 06:46:47, nothing until manual restart at 06:54:41 ✗

Environment

• Claude Code: 2.1.37
• OS: Windows 10 (19045.6466)
• Terminal: Windows Terminal + PowerShell
• Working directory: D:\ (drive root, ~190K files)

Workaround

Run from a subdirectory instead of a drive root.

Notes

• .claudeignore does not help because the scan uses --no-ignore
• The $RECYCLE.BIN and System Volume Information directories produce access denied errors during the scan
• This also affects the session resume picker (claude --resume) — arrow keys and typing don't work