## Summary

When the session is locked around suspend/resume, `omarchy-shell` crashes while creating the **lock surface**, leaving an orphaned `ext-session-lock` in Hyprland. The screen shows a blank/frozen "error page" with no UI, then ~8s later the shell relaunches, detects the stranded lock (`lock-stranded: recovering`), and only then shows the password prompt.

## Observed behavior

- User sees an error/blank page load, then the lock screen appears afterwards (~8 s gap).
- This recurs: the shell crashed **4 times today** (09:14, 09:19, 10:06, 12:09) with the same failure.

## Expected behavior

Locking during/around suspend should not kill the shell; the lock surface should be created for the active output and the password screen should appear immediately.

## Root cause sequence (from journal)

```
12:09:16  systemd[765]: omarchy-sleep-lock.service: Scheduled restart job, restart counter is at 2.
12:09:17  systemd[765]: Started Lock Omarchy before suspend.
12:09:18  omarchy-shell[286934]: DEBUG: Not creating lock surface for screen QScreen(0x..., name="") as it is not backed by a valid wayland output.
12:09:22  omarchy-shell[286934]: WARN qt.qpa.wayland: Could not create EGL surface (EGL error 0x3000)
12:09:22  omarchy-shell[286934]: WARN qt.qpa.wayland: eglSwapBuffers failed with 0x300d, surface: 0x0
12:09:22  omarchy-shell[286934]: WARN: The Wayland connection experienced a fatal error: Invalid argument
12:09:22  omarchy-shell[293027]: Omarchy shell exited with status 255; relaunching.
12:09:23  omarchy-shell[293200]: Launching config ...
12:09:24  omarchy lock 2026-09-10T04:09:24.768Z lock-stranded: recovering
12:09:24  omarchy lock ... lock-requested
12:09:24  omarchy lock ... lock-pending: screen-stabilizing
12:09:25  omarchy lock ... secure=true            <- password screen visible
12:09:29  PAM authenticated, unlocked
```

Other recurring notes at the same time:
- `Got removal for monitor "FALLBACK" which was not previously tracked` (quickshell.hyprland.ipc)
- Concurrent `ddcutil` processes failing to `flock()` `/dev/i2c-7` (`Max wait time 3000 milliseconds exceeded after 32 flock() calls`, call stack `i2c_check_bus`) at the lock moment, affecting DDC brightness blank/restore.

## Steps to reproduce

1. Suspend the machine (or lock while outputs are mid-reconfigure, e.g. external monitor connect/disconnect).
2. `omarchy-sleep-lock` triggers the shell to lock before outputs are stably back.
3. Shell attempts EGL lock surface on a `QScreen(name="")` not backed by a valid Wayland output → fatal Wayland error → crash → orphaned lock → blank screen until relaunch recovers.

## System details

- Omarchy: **4.0.3-1** (ID `omarchy`, BUILD_ID 4.0.3)
- Hyprland: **0.56.2** (commit efb5099, tag v0.56.2)
- quickshell: **0.3.1-1**
- CPU: Intel Celeron N5095 (Jasper Lake)
- GPU: Intel UHD Graphics (i915) — Wayland session
- Kernel/driver: `i915`

## Diagnostics

Journal (grep for `egl|EGL|fatal error|exited with status|omarchy_sleep-lock`):
```
journalctl --since "2026-09-10 00:00" --no-pager | grep -E "exited with status|fatal error|EGL"
```
Quickshell QML logs confirm the stranded-lock recovery path (`omarchy lock ... lock-stranded: recovering`) in `/run/user/1000/quickshell/by-id/*/log.qslog`.

## Suggested fix direction

In `shell/plugins/lock/Service.qml`, avoid creating the lock surface while no output is valid (already guarded by `hasRealScreen()` for the lock request, but the `WlSessionLockSurface` creation can still race output teardown during suspend), and/or make the lock-surface creation resilient so an EGL failure cannot take down the whole shell right when the machine is suspending.