### What happened

`omarchy update` finished its package transaction, then its shell restart step failed:

```
Restarting shell
All plugins have been reloaded
Omarchy shell did not become ready after restart.
```

The session was left with **no bar and no notifications**, and the message gives no hint as to why. `omarchy restart shell` reproduced the same failure on every subsequent attempt.

### Root cause

`bin/omarchy-restart-shell` only derives the Hyprland instance signature when the variable is **unset**, so a variable that is set but **stale** is trusted blindly:

https://github.com/basecamp/omarchy/blob/master/bin/omarchy-restart-shell#L14-L19

```bash
# Allow running from outside the session (e.g. over ssh) by deriving the
# Hyprland instance signature from the newest instance runtime dir.
if [[ -z ${HYPRLAND_INSTANCE_SIGNATURE:-} ]]; then
  hypr_dir=$(find "${XDG_RUNTIME_DIR:-/run/user/$UID}/hypr" -mindepth 1 -maxdepth 1 -type d -printf '%T@ %p\n' | sort -n | tail -n 1 | cut -d' ' -f2-)
  [[ -n $hypr_dir ]] && export HYPRLAND_INSTANCE_SIGNATURE=${hypr_dir##*/}
fi
```

If Hyprland has been restarted since a terminal was opened, that terminal still exports the **previous** instance signature. The script keeps it, and the `hyprctl dispatch` that spawns the shell then talks to a dead socket:

```
$ hyprctl version
Couldn't connect to /run/user/1000/hypr/<old-signature>/.socket.sock. (4)
```

That dispatch's stderr is discarded, so the failure is silent; the script polls 20 times and reports only the generic "did not become ready" message.

Notably, line 8 already solves exactly this class of staleness for a different variable, with a comment describing the same reasoning:

```bash
session_omarchy_path=$(systemctl --user show-environment 2>/dev/null | sed -n 's/^OMARCHY_PATH=//p' | tail -n 1)
```

The systemd user environment holds the **correct, live** signature — Hyprland updates it on session start — so the same treatment would have avoided this. On the affected machine:

```
$ systemctl --user show-environment | grep HYPRLAND_INSTANCE_SIGNATURE
HYPRLAND_INSTANCE_SIGNATURE=efb5099...1788366025_1900155456   # live

$ echo $HYPRLAND_INSTANCE_SIGNATURE                            # in the terminal
efb5099...1788257371_1303222695                                # dead, ~30h older
```

Two instance dirs were present in `/run/user/1000/hypr/`; only the newer one had a responding socket.

### Steps to reproduce

1. Open a terminal.
2. Restart Hyprland (or log out and back in), leaving that terminal open.
3. In that same terminal, run `omarchy restart shell` (or `omarchy update`, which calls it).
4. It prints "Omarchy shell did not become ready after restart" and the session is left without a shell.

### Expected

The shell restarts. A signature that no longer has a live socket should be discarded and re-derived, rather than trusted because it happens to be non-empty.

### Suggested fix

Prefer the systemd user environment, and/or validate the socket before trusting the variable — e.g. treat a signature whose `.socket.sock` does not respond the same as an unset one, falling through to the existing newest-runtime-dir logic. Surfacing the `hyprctl dispatch` error would also turn a mystifying failure into a diagnosable one.

### Workaround

Re-exporting the live signature, or launching `omarchy-launch-shell` directly, brings the shell back:

```bash
export HYPRLAND_INSTANCE_SIGNATURE=$(systemctl --user show-environment | sed -n 's/^HYPRLAND_INSTANCE_SIGNATURE=//p')
omarchy restart shell
```

### System details

- Omarchy 4.0.2-1 (`bin/omarchy-restart-shell` on `master` is byte-identical to the installed copy)
- Hyprland via uwsm, quickshell 0.3.1-1
- Linux 7.1.9-arch1-2
- Intel Core i7-7700HQ; Intel HD Graphics 630 + NVIDIA GTX 1050 Ti Mobile (`nvidia-580xx-dkms`)