Linux: clean-exit renderer dialog during shutdown, followed by main-process SIGTRAP

## Observed behavior

On Omarchy/Arch Linux with Hyprland, VS Code became stuck on an error dialog during a development-environment setup session involving extension installation. The user reported that the desktop was stuck on this screen.

The supplied screenshot shows:

> The window terminated unexpectedly (reason: 'clean-exit', code: '0')

A second, dimmed dialog is visible behind it:

> An application Visual Studio Code is not responding.

The foreground dialog offers Close / Reopen; the background dialog offers Terminate / Wait.

## Expected behavior

An orderly shutdown should not leave an unexpected-renderer-exit dialog or crash the main process. If termination is abnormal, the application should remain recoverable without getting stuck between overlapping failure dialogs.

## Incident sequence (not a deterministic reproduction)

1. Install VS Code and development extensions on Omarchy/Arch Linux.
2. Open a local project with `code --verbose --new-window `.
3. During the setup session, encounter the dialogs described above. Logs establish that a shutdown began immediately before the renderer exit; the action or signal sender initiating shutdown is unknown.

This incident has not been deliberately reproduced. Extension-disabled, GPU-disabled, and Insiders comparisons have not been performed. Installing extensions is the surrounding activity, not an established cause.

## Correlated evidence

All timestamps below are 2026-09-15, CEST (UTC+02:00), from the same local session.

Sanitized excerpts from `main.log`:

```text
08:02:21.859 [trace] Lifecycle#app.on(before-quit)
08:02:21.859 [trace] Lifecycle#onBeforeShutdown.fire()
08:02:21.859 [trace] Lifecycle#window.on('close') - window ID 1
08:02:21.859 [trace] Lifecycle#unload() - window ID 1
08:02:21.863 [info] [network process] gone - pid: 709367, exitCode: 15, crashed: true, crashedPreIPC: false
08:02:21.871 [error] AgentHostProcessManager: agent host terminated unexpectedly with code 15
08:02:21.875 [info] AgentHostProcessManager: agent host started
08:02:21.876 [info] [network process] launched with pid 710011
08:02:21.879 [info] Extension host with pid 709509 exited with code: 15, signal: unknown.
08:02:21.879 [error] [UtilityProcess id: 1, type: extensionHost, pid: 709509]: crashed with code 15 and reason 'killed'
08:02:21.880 [error] [UtilityProcessWorker]: terminated unexpectedly with code 15, signal: unknown
08:02:21.881 [error] [UtilityProcess type: shared-process, pid: 709458]: crashed with code 15 and reason 'killed'
08:02:21.884 [error] CodeWindow: renderer process gone (reason: clean-exit, code: )
```

At **08:02:26**, systemd-coredump recorded the main process `/usr/share/code/code` (PID 709323) terminating with **SIGTRAP**, `si_code: SI_KERNEL`. The command line was `/usr/share/code/code --verbose --new-window `.

The stored core is explicitly marked **truncated**. GDB enumerated 55 threads, but backtraces stopped at frame 0 with inaccessible stack memory. The main-thread address was `0x000056380339b7e1`, unresolved; no native function or library is attributed as the cause.

Additional checks:

- No OOM-kill or NVIDIA Xid/GPU-fault entries found in the current boot's kernel journal.
- After the incident, memory availability was approximately 43 GiB out of 62 GiB, with 242 MiB swap used.
- The coredump-processing service itself recorded 28.4 GiB peak memory and about 49 seconds wall time, ending at 08:03:16. Whether this contributed to the perceived desktop stall is unknown.
- Hyprland's client-query IPC responded during subsequent diagnosis; no failed user systemd units were listed. These later checks do not disprove the user's earlier desktop freeze.
- Package integrity check: `visual-studio-code-bin: 3067 total files, 0 altered files`.

**Inference, not established cause:** near-simultaneous code-15 helper exits and the shutdown lifecycle suggest a termination/shutdown interaction. The sender and reason are unknown. No extension, GPU driver, or Omarchy component has been established as responsible.

## Environment

- VS Code 1.137.0, stable; commit `645f29cc3176500b4b5762ba887cf2a7f0ffdf2c`
- Arch package `visual-studio-code-bin 1.137.0-1`, installed that morning
- Omarchy `4.0.3-1`
- Hyprland `0.56.2-2`
- Linux `7.2.3-arch1-3`, x86_64
- Intel Core i9-10900K; NVIDIA GeForce GTX 1660 SUPER
- NVIDIA userspace driver `610.57.04-1`
- No `~/.config/code-flags.conf` was present when inspected

Installed extensions at diagnosis:

```text
ms-python.vscode-python-envs 1.36.0
ms-python.debugpy 2026.6.0
ms-python.vscode-pylance 2026.3.1
ms-python.python 2026.4.0
rust-lang.rust-analyzer 0.3.3049
vadimcn.vscode-lldb 1.12.3
dbaeumer.vscode-eslint 3.0.34
esbenp.prettier-vscode 12.4.0
```

## Related reports checked

#326092 describes a 1.129.0 startup crash/relaunch loop. This report instead has a recorded before-quit sequence followed by a main-process SIGTRAP on 1.137.0; equivalence is not established. #329582 and #323874 concern macOS clean-exit/code-1 incidents. No confirmed matching report was found in the searches performed.

Only sanitized excerpts are included. The core dump and full diagnostic logs are not uploaded because they may contain private data. The screenshot text is transcribed above; the image is not attached.

Filed by GPT-6 Astra via Oh My Pi, at the user's request.