Persistent "API Error: Connection error" and Request Timed Out in WSL2 Environment

Claude Code frequently encounters "API Error: Connection error" and "Request timed out" during normal operations in WSL2. This issue persists across different accounts and different machines, often leading to 5-10 minute delays or complete failure of tasks.

Error Logs

● Read(file_path)
  ⎿  API Error: Connection error.
✻ Cogitated for 9m 22s

❯ continue 
⎿ Request timed out. Check your internet connection and proxy settings
 Retrying in 27 seconds… (attempt 7/10)

🔍 Environment Details

• OS: Windows 10 Pro N (Build 19045) / WSL2
• Kernel: 6.6.87.2-microsoft-standard-WSL2
• Docker: Docker Desktop with WSL2 integration enabled
• Network: NAT mode (default) / Attempted Mirrored mode

🕵️ Diagnosis & Observations

1. Port Conflicts: There appears to be a relationship between running Docker containers and the frequency of these API errors.
2. Resource Exhaustion: During these errors, CPU usage for the claude process often spikes to 70-110%.
3. WSL2 Networking: The NAT boundary in WSL2 seems to interfere with the OAuth callback and long-lived API connections.
4. Leaked Processes: We have observed cases where dozens of MCP processes remain active, exhausting system resources.

💡 Requested Action

We request a thorough investigation into how Claude Code handles network interruptions in WSL2 and a more robust retry mechanism that doesn't hang the process for several minutes. Additionally, clarification on the impact of local port usage (e.g., Docker) on the API's connection stability is needed.

---
Evidence attached: (Logs and environment snapshots)