## Summary

`omarchy-system-shutdown` schedules `systemctl poweroff` via `nohup bash -c "sleep 2 && systemctl poweroff --no-wall" >/dev/null 2>&1 &`. Although `nohup` and `&` detach from the controlling terminal, the process **stays inside the user's `session-N.scope`**. When systemd begins the shutdown sequence it stops that scope first, killing the very `systemctl poweroff` it depends on. Result: every shutdown is "dirty".

## Steps to reproduce

1. Open the Omarchy menu → Power → Shutdown
2. After reboot inspect the previous boot:
   ```bash
   journalctl -b -1 --no-pager | grep -E "Stopping timed out|Killing process .* systemctl|Failed unmounting"
   ```

## Expected

Clean shutdown, all mounts unmounted, no `session-N.scope: Failed with result 'timeout'`.

## Actual (consistent across 5+ boots on my system)

```
systemd[1]: session-2.scope: Stopping timed out. Killing.
systemd[1]: session-2.scope: Killing process <PID> (systemctl) with signal SIGKILL.
systemd[1]: session-2.scope: Failed with result 'timeout'.
systemd[1]: Failed unmounting /var/cache.
umount[<PID>]: umount: /var/cache: target is busy.
```

The `Failed unmounting /var/cache` follow-on happens because background services (plocate-updatedb, fwupd, pacman temp dirs) are still holding `/var/cache` when the kill cascade reaches the unmount step — `omarchy-hyprland-window-close-all` does not give them time to settle, and the SIGKILL of the orchestrating systemctl truncates the orderly shutdown sequence.

## Environment

- Omarchy 3.8.0
- Arch / CachyOS base, systemd 257.x, btrfs root with `@cache` subvol mounted at `/var/cache`
- Hyprland 0.55.0 under uwsm

## Root cause

```bash
nohup bash -c "sleep 2 && systemctl poweroff --no-wall" >/dev/null 2>&1 &
```

`nohup` only blocks SIGHUP. The detached background process inherits the systemd unit (`session-N.scope`). When `systemd-logind` tells systemd to stop the session, that scope is terminated with the standard stop sequence — `SIGTERM`, wait `TimeoutStopSec`, then `SIGKILL`. The pending `sleep 2 && systemctl poweroff` is killed before it can complete its job, leaving the actual shutdown to fall back to systemd's own mechanism (which is what currently powers the machine off, but only after the noisy timeout).

This is masked but not fixed by lowering `DefaultTimeoutStopSec` (#3443 / #1886 / #5133): the cleanup is still violent, just faster.

## Proposed fix

Run `poweroff` in a transient systemd unit so it is owned by PID 1, not by the session scope:

```bash
systemd-run --no-block --collect --unit omarchy-poweroff /usr/bin/systemctl poweroff --no-wall
```

Or simpler — drop the `sleep`/detach trick entirely: `omarchy-hyprland-window-close-all` already runs synchronously and `systemctl poweroff` returns immediately after handing off to PID 1:

```bash
omarchy-hyprland-window-close-all
sleep 1
exec systemctl poweroff --no-wall
```

I'm happy to send a PR with whichever pattern you prefer.