### System details

- ThinkPad, 11th Gen Intel i7-1165G7, i915 Iris Xe (Tiger Lake)
- Omarchy 4.0.0-1, Hyprland 0.56.2 (`efb50993780079460b0cbed1363e2166a2de1d9f`), kernel 7.1.8-arch1-3, uwsm
- Internal: eDP-1 AU Optronics 1920x1200@60
- External: HDMI-A-1 BenQ RL2455 1920x1080@60

`omarchy-debug --no-sudo --print` hung on this machine (no output after well over a minute) so it is not attached. Journal excerpt is below.

### What's wrong?

Plugging in an HDMI monitor and then closing the lid a couple of seconds later kills **every GUI client**. Hyprland itself stays up. The session looks alive (bar may relaunch) but terminals, Nautilus, Brave, portals, fcitx5, and XWayland are gone.

This is a Hyprland 0.56.2 compositor bug (`wl_output` global destroyed while clients are still binding it). Omarchy’s clamshell helper and monitor watcher are what fire that path on a stock laptop: they write a specific-output `disabled = true` rule for `eDP-1` and `hyprctl reload` on hotplug / lid close.

### Steps to reproduce

Stock catch-all monitor config (no `eDP-1` rule in `~/.config/hypr/monitors.lua`):

```lua
hl.monitor({ output = "", mode = "preferred", position = "auto", scale = omarchy_monitor_scale })
```

1. Laptop lid open, no external display.
2. Plug in HDMI.
3. Close the lid a couple of seconds later.
4. Watch the GUI session empty out. `hyprctl clients` has almost nothing; `ps` may still show Brave.

### What Omarchy does in that window

On `monitoradded`, `omarchy-hyprland-monitor-watch` calls `omarchy-hyprland-monitor-clamshell` and also `hyprctl reload` (0×0 EDID recovery) with retries at 1s / 3s / 7s.

When the lid is then closed (`omarchy-hw-clamshell` && external active), `omarchy-hyprland-monitor-clamshell` writes:

```lua
-- ~/.local/state/omarchy/toggles/hypr/internal-monitor-clamshell.lua
hl.monitor({ output = "eDP-1", disabled = true })
```

and runs `hyprctl reload` again. That toggle file is loaded as Hyprland config (`require("default.hypr.toggles")`).

So even with a catch-all-only `monitors.lua`, Hyprland sees a specific-output disable of `eDP-1` plus one or more reloads while clients are still binding the new HDMI `wl_output`.

### Journal (2026-08-23 08:02:26 local)

Hyprland pid 1201 had been running since 2026-08-16. It did not crash.

```
uwsm_hyprland.desktop[1201]: error in client communication (pid 1268)
omarchy-shell[1268]: WARN: The Wayland connection experienced a fatal error: Protocol error
xdg-desktop-portal-hyprland[1549]: wl_registry#2: error 0: global wl_output (79) is unavailable
fcitx5[1401]: wl_registry#2: error 0: global wl_output (79) is unavailable
xdg-desktop-portal-gtk[1522]: Error 71 (Protocol error) dispatching to Wayland display.
XWAYLAND: wl_registry#2: error 0: global wl_output (79) is unavailable
XWAYLAND: (EE) failed to dispatch Wayland events: Protocol error
```

Same second: omarchy-shell, both portals, fcitx5, XWayland, terminals, Nautilus, Brave all gone. `hyprctl clients` afterwards showed only a newly opened terminal.

### Aftermath (XWayland / Brave)

XWayland is never respawned. Hyprland keeps listening on `/tmp/.X11-unix/X0`, so `xprop -root` hangs. Brave Origin runs (`ozone-platform=wayland`) with no Hyprland window: Chromium startup is blocked in `xdg-settings` → `xprop -root _DT_SAVE_MODE`. Logout/login is required to get X11 back.

### Upstream

This is the Hyprland registry bind race. The living thread is [hyprwm/Hyprland#12996](https://github.com/hyprwm/Hyprland/discussions/12996); the matching issue [#15635](https://github.com/hyprwm/Hyprland/issues/15635) was auto-closed. This machine’s report is [this comment](https://github.com/hyprwm/Hyprland/discussions/12996#discussioncomment-18121477).

On 0.56.2, a specific-output rule for the panel being disabled is what turns “one client dies” into a session wipe. Omarchy writes that rule itself.

### Suggested Omarchy-side change

Until Hyprland stops fataling clients on output removal:

- Do not persist `hl.monitor({ output = "<internal>", disabled = true })` and `hyprctl reload` on lid close / HDMI add.
- Blank the internal panel without destroying the `wl_output` (backlight 0 works here; Hyprland DPMS on eDP-1 does not stick because the bar still lives on that output).
- Move occupied workspaces to the external monitor.
- Do not `hyprctl reload` on every `monitoradded` / `monitorremoved`. Keep 0×0 EDID recovery for actual modeless outputs, not for a normal HDMI plug.

### Workaround in use on this machine (not a fix)

User wrappers in `~/.local/bin` shadow the packaged clamshell/watch scripts: no disable-rule, no reload on HDMI add, backlight 0 + workspace move. Revert is `~/.local/state/omarchy/backups/20260823-clamshell-mitigation/revert.sh`. This does not fix a raw Hyprland HPD bounce.


*This was composed and posted by Grok Build at the direction of this user.*