[BUG] Claude Code CLI automated reconnect loop triggers false-positive account ban with zero appeal recourse

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 running Claude Code CLI in an environment with intermittent network packet loss or unstable Wi-Fi/ISP jitter, the CLI client enters an aggressive automated reconnection loop.

Instead of gracefully backing off, the rapid burst of failed TLS handshakes and reconnect attempts generated by the CLI client causes Anthropic's backend security gateways / WAF to misclassify the legitimate developer session as "suspicious automated signals / bot abuse".

This immediately triggers an irrevocable account suspension under "Usage Policy", even though the user is engaged in standard, legitimate engineering work (in my case, embedded algorithm development on a local Mac mini).

To make matters worse, there is a severe UX/procedural failure: once the account is flagged, the web appeals portal at https://claude.ai/restricted permanently deadlocks into a static "Your account will remain closed" page with ZERO active form inputs or appeal mechanisms. Support email inboxes (usersafety@ and support@) are entirely handled by automated bots that auto-close tickets, leaving developers with zero recourse to provide technical logs or report the client-side retry bug.

What Should Happen?

1. Claude Code CLI should implement proper exponential backoff and jitter algorithms during connection drops, rather than generating a high-frequency burst of reconnect requests that triggers backend abuse detection.
2. Server-side security classifiers should differentiate between client-side TCP/TLS reconnection retries from an established developer CLI and actual malicious scraping/botnet activity.
3. The appeals portal should not dead-end into a permanent lockout when client-side infrastructure errors occur.

Error Messages/Logs

- Account Status: "An internal investigation of suspicious signals associated with your account indicates a violation of our Usage Policy."
- Suspension Reference: TS-01a0f49c-1cf1-7256-86d7-52fb0859c71d
- Appeal Notice Reference: TS-01a0feee-8e61-7575-9449-095db0d51d59
- Affected Account Email: chendongtse@gmail.com

Steps to Reproduce

1. Launch Claude Code CLI in a local development directory.
2. Introduce local network packet loss or intermittent connectivity drops during an active session.
3. Observe Claude Code CLI rapidly retrying connections and failing TLS handshakes in quick succession.
4. Shortly after, the backend terminates the account for "suspicious signals associated with your account" (Usage Policy violation).
5. Attempting to access https://claude.ai/restricted reveals an inactive static page with no way to appeal.

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

v2.1.289

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_