## What happened

Super+Space has repeatedly stopped showing the Omarchy menu during an existing desktop session. The user binding remains configured. Rebinding it from the global shortcut to `omarchy-menu toggle root` did not prevent the recurrence.

On 2026-10-06, a local recovery wrapper attempted to open the menu, did not detect a visible `omarchy-menu` overlay, restarted the Omarchy shell, and then detected the menu on screen. This is a confirmed intermittent menu failure/recovery; the underlying cause has not been isolated.

## Expected behavior

Super+Space should continue opening and closing the menu throughout the session, including after display output reconfiguration, without requiring a shell restart.

## Observed sequence and reproduction status

1. Start a normal Omarchy desktop session with the configuration described below.
2. Later in the session, press Super+Space. Intermittently, the menu does not appear.
3. Restart the shell and request the same menu again; the menu appears.

The exact trigger is not yet reproducible on demand. The earlier shell log records two transitions through a zero-output / placeholder-screen state in the same session, making display output reconfiguration a candidate trigger. We have not established causality or reproduced this with third-party plugins disabled.

## Evidence

Times below are local UTC−05:00 on 2026-10-06.

The original Quickshell instance logged:

```text
09:29:48.992 INFO qt.qpa.wayland: There are no outputs - creating placeholder screen
09:29:49.057 WARN scene: Handler was registered but will not be used because another handler is registered for target sid.sessions
09:29:49.800 WARN: Layershell screen does not correspond to a real screen. Letting the compositor pick.
09:52:02.357 INFO qt.qpa.wayland: There are no outputs - creating placeholder screen
09:52:02.942 WARN: Layershell screen does not correspond to a real screen. Letting the compositor pick.
09:52:05.257 WARN quickshell.hyprland.ipc: Got removal for monitor "FALLBACK" which was not previously tracked.
11:49:46.826 INFO: Exiting due to IPC request.
```

The local recovery log recorded:

```text
2026-10-06T11:49:46-05:00 Menu did not become open; restarting Omarchy shell.
omarchy-shell is not responding
omarchy-shell is not responding
2026-10-06T11:49:50-05:00 Recovered; menu visible.
```

The two IPC errors above occur **after recovery starts** and may be readiness probes against the replacement shell. They do not establish that the original shell was unresponsive. The original instance answered the IPC exit request. At this occurrence, per-call exit status/timing and the complete pre-restart layer state were not yet recorded.

## System and configuration

| Component | Version / details |
|---|---|
| Omarchy package | `omarchy-dev 4.0.0.r6720.g8e02fc8-1` |
| Packaged version file | `4.0.0.alpha` |
| Kernel | `7.2.5-4-omarchy` |
| Hyprland | `0.56.2-4` |
| Quickshell | `0.3.1-1` |
| Qt base / declarative | `6.11.2-3` / `6.11.2-2` |
| CPU | Intel Core i5-9500T |
| GPU | Intel UHD Graphics 630, i915 |
| Display configuration | Preferred mode, scale 2; output `DP-2` in session logs |

This is a customized installation. The built-in menu remains in use; third-party bar widgets/plugins include sid.sessions, Omacorners, FreeBuds, Startup Map, Vision Hub, crypto widgets, a shutdown timer and Proton VPN. A stock-plugin reproduction is still pending.

## Local mitigation and next evidence

A user-owned wrapper now verifies an overlay menu surface (rather than mere presence of the parked 1×1 surface), resets the menu's internal open state before `summon root`, and performs one bounded shell restart if opening fails. It is a mitigation, not a proposed root-cause fix. Its initial IPC calls use a 0.7-second timeout, which is another variable to control when reproducing.

Before any future recovery, the wrapper will save the original keypress monitor/layer observation, live monitor/layer/binding/config state, shell/menu/direct-Quickshell IPC responses with timing, plugin state, processes, runtime disk usage, journal excerpts and Quickshell logs. A second capture follows recovery. Screenshots are saved locally when an unlocked real output can be confirmed. Nothing from these bundles is uploaded automatically; selected evidence will be reviewed and added to this issue.

No screenshot of the original failure was preserved. The automated recorder was installed afterward. The collection environment could read existing log files but could not directly query the desktop IPC; those failed agent-side queries are not being presented as evidence that the desktop was broken.

Related: #9745 has similar placeholder-screen warnings followed by unusable menus, but describes a stranded-lock crash loop and full runtime filesystem. We have not observed or established those mechanisms here and are not assuming the same root cause.