[BUG] Silent session exits with resume link — persists across multiple machines after Feb 10 incident

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?

Summary

Claude Code sessions silently exit mid-conversation with only a resume link, no error message.
This began the evening of Feb 10, 2026 and continues on Feb 11 despite the status page showing
all services operational. Affects multiple machines on the same account simultaneously.

Environment

• Claude Code version: 2.1.39 (also reproduced on 2.1.19)
• OS: macOS 15 (Darwin 25.2.0, arm64) and Ubuntu (Linode)
• Account: Max subscription, under 10% usage allowance
• Machines affected: Mac Mini (macOS), Linode (Ubuntu), MacBook (macOS — intermittent)

Reproduction

1. Run claude (no flags) in any directory, including bare /tmp/test-claude with no CLAUDE.md or

project config

1. Session starts, may show a permission prompt or begin responding
2. Session silently exits within seconds, displaying only:

Resume this session with:
claude --resume <session-id>

1. No error message, no stack trace, no crash log
2. Resuming with the provided command also exits quickly

What I've ruled out

• Not flag-related: Happens with plain claude and with --dangerously-skip-permissions
• Not project/config-related: Happens in bare /tmp directory with no CLAUDE.md
• Not version-related: Happens on 2.1.39 and 2.1.19
• Not usage-limited: Account is under 10% of Max plan allowance
• Not MCP-related: Removed all MCP config, same behavior. One MacBook session that connected

successfully uses MCP without issues

• Not machine-specific: Reproduced on 3 separate machines (2 macOS, 1 Linux)

Key observation

One session on a MacBook that connected successfully has remained stable for 30+ minutes through
heavy parallel tool use, subagent spawning, web searches, and MCP operations. The issue appears
to affect new connection establishment, not existing sessions. This suggests backend
instability where some nodes are healthy and others are not.

Timeline

• Feb 9 evening: First noticed silent exits across all machines
• Feb 10 ~02:05 PST: Anthropic status page showed Opus 4.6 Fast Mode errors (~5 hour warning

state)

• Feb 11 morning: Status page shows "operational" but new sessions on Mac Mini and Linode still

exit immediately. Only one MacBook session has remained stable.

What Should Happen?

Expected behavior

Sessions should either connect reliably or show a meaningful error message explaining the
failure (e.g., "service temporarily unavailable, retrying...") rather than silently exiting.

Error Messages/Logs

Steps to Reproduce

run claude
interact 30-90 seconds

Claude Model

Not sure / Multiple models

Is this a regression?

Yes, this worked in a previous version

Last Working Version

_No response_

Claude Code Version

2.1.19 & 2.1.39

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_