[BUG] Session silently terminates while churning — no error message, only "Resume this session with:"

What's Wrong?

During normal operation, Claude Code sometimes silently terminates a session while actively processing ("Churning..." state). The session drops with no error message explaining what went wrong. The only output is:

Resume this session with:
claude --resume 687ee572-6d74-4b18-a1b5-9776b47...

There is no indication of why the session was interrupted — whether it hit a context limit, a token ceiling, a timeout, an API error, or something else entirely. The user is left guessing.

What Should Happen?

When a session is interrupted, Claude Code should display a clear error message explaining the reason before showing the resume command. For example:

• "Session interrupted: context window limit reached"
• "Session interrupted: API timeout after 120s"
• "Session interrupted: rate limit exceeded, retry in Xs"

Steps to Reproduce

1. Use Claude Code in VS Code integrated terminal with a Max plan
2. Work on a task that involves multiple tool calls and file reads (building up context)
3. At some point the session enters "Churning..." state
4. Session silently drops — shows only the resume session ID with no error explanation

This is intermittent and hard to reproduce on demand, but happens frequently enough to be disruptive.

Error Messages/Logs

No error message is displayed — that's the bug. The only output before the session drops:

* Churning... (1m 13s · ↑ 1.7k tokens · thought f...

>

  esc to interrupt

Resume this session with:
claude --resume 687ee572-6d74-4b18-a1b5-9776b47...

Environment

• Claude Code Version: 2.1.85
• Claude Model: Opus 4.6 (1M context)
• Platform: Max plan
• Operating System: macOS 26.3.1 (Tahoe)
• Terminal/Shell: VS Code integrated terminal (zsh)
• Machine: Mac Studio, Apple M2 Ultra, 128 GB
• VS Code: 1.112.0
• Claude Code Extension: 2.1.85

Is this a regression?

Don't know — this behavior has been present for as long as I've been using Claude Code, but it may have become more frequent in recent versions.

Additional Information

Screenshot attached showing the silent drop. The session was actively churning (1m 13s, 1.7k tokens into a thought) when it terminated. No crash dialog, no error code, no explanation — just the resume command.

This is especially frustrating because:

1. Without knowing the cause, users can't avoid triggering it
2. The resume command doesn't always successfully restore the session state
3. It breaks flow during complex multi-step tasks