Sandbox VM repeatedly reboots in Claude Code web session — blocking long-running job

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?

The sandbox VM/container for my Claude Code web session repeatedly reboots on its own (not an OOM kill — uptime resets to "up 0 min" and dmesg shows a fresh boot sequence each time). This has happened more than 15 times over ~30 hours (2026-10-11 to 2026-10-12), with reboot intervals ranging from ~5 minutes to ~2 hours.

Session ID: session_01HWVoLm2ZC9m6W6UgwJ3HGG

I'm running a background nbclient process executing a 235-cell Jupyter notebook (expected ~5-7 hours uninterrupted, based on earlier successful runs in the same session). Ruled out: no OOM entries in dmesg, memory healthy (~5GB/15GB used), disk healthy (~25GB free), no orphaned processes, code verified correct (no Python Tracebacks — only the VM reboots from underneath it).

Across the last 7 consecutive relaunch attempts, the job died within 5-15 minutes each time, never reaching the computationally heavy stage of the pipeline.

What Should Happen?

The sandbox VM should remain stable for the duration of a long-running background compute job, or I should be able to request a replacement/reset of the VM for this session if it is unhealthy.

Error Messages/Logs

dmesg after each reboot shows (example, repeats each incident):
[0.000000] Command line: ... --firecracker-init ...
[1.x] virtio_blk ... new size ...
[3.x] EXT4-fs (vda): mounted filesystem ... r/w

uptime immediately after each incident: "up 0 min" (or "up 1 min" / "up 2 min")

No OOM-killer entries found in dmesg at any point.

Steps to Reproduce

1. Start a Claude Code web session and launch a long-running background job (e.g. a detached Python process executing a multi-hour Jupyter notebook via nbclient).
2. Let it run unattended for an extended period.
3. Periodically check uptime and dmesg in the sandbox.
4. Observe that the VM reboots on its own at irregular intervals (minutes to ~2 hours apart), confirmed by uptime resetting to near-zero and dmesg showing a fresh boot sequence — not a process-level crash.
5. The background job is killed each time, requiring a manual relaunch.

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

N/A - not applicable, this is not a version regression

Claude Code Version

N/A - Claude Code on the web (cloud session), not a local CLI install

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

Note: this issue is specifically about the sandbox VM for a Claude Code WEB/cloud session (not a local CLI install) — the OS/Terminal fields above don't strictly apply since there's no local terminal involved. The instability is in Anthropic's remote sandbox infrastructure itself.