### System details

AZW SER3 mini PC, AMD Ryzen 5 3500U with Radeon Vega (amdgpu), single external monitor ASUS PA278QV on HDMI-A-1, Omarchy 4.0.4-1 (stable channel), Hyprland 0.56.2-2, aquamarine 0.15.0-2, quickshell 0.3.1-1, kernel 7.2.5-3-omarchy.

### What's wrong?

Locking the screen (manually or by idle) puts the shell into a crash/relaunch loop that power-cycles the monitor roughly every 28 seconds for as long as the session stays locked. In one lock today it ran for 71 minutes and relaunched omarchy-shell 89 times. The same loop preceded a lock on 2026-09-25 (40 relaunches) and three locks on 2026-09-10/11.

**Sequence, from the journal and Hyprland's aquamarine log (every cycle is identical):**

1. Lock engages: `omarchy lock ... secure=true`.
2. The lock plugin's blank timer runs `omarchy-brightness-display off`, which dispatches `hl.dsp.dpms disable`. aquamarine logs `drm: Disabling output HDMI-A-1`.
3. About 20 s after the lock, the kernel reports the connector gone: `drm: Got a hotplug event` then `drm: Connector HDMI-A-1 disconnected`. Hyprland removes the only output and creates the FALLBACK monitor.
4. omarchy-shell logs `qt.qpa.wayland: There are no outputs - creating placeholder screen`, Hyprland prints `error in client communication (pid <shell>)`, and the shell dies:
   ```
   WARN quickshell.hyprland.ipc: Got removal for monitor "FALLBACK" which was not previously tracked.
   WARN: The Wayland connection experienced a fatal error: Invalid argument
   Omarchy shell exited with status 255; relaunching.
   ```
5. The relaunched shell re-locks (`lock-stranded: recovering` -> `lock-requested` -> `secure=true`), the monitor reconnects (`Connector HDMI-A-1 connected`, modeset [email protected]), and the blank timer fires again. Back to step 2.

**Why it matters beyond the flashing monitor:** every reconnect makes Hyprland re-send `zwp_linux_dmabuf_feedback_v1` to all clients. With gtk4 4.22.4 (what the stable mirror still serves) that hits a GTK use-after-munmap bug (GTK MR !10166, fixed in 4.22.5), and the ghostty single-instance daemon segfaulted on 5 separate locks (2026-09-10 x2, 09-11, 09-25, 10-06), taking all terminals with it. The segfault itself is GTK's bug and is fixed upstream, but the stable mirror is still on 4.22.4 and the lock loop is what triggers it hundreds of times.

**Expected:** after the display is blanked, the shell should survive the output disappearing (or Hyprland should not post a protocol error to it), and the lock should not keep re-blanking a monitor that drops its HDMI link on DPMS off.

**Steps to reproduce (on this hardware):** lock the screen with the PA278QV as the only monitor and wait ~25 s. Watch `journalctl -f` for `There are no outputs`, `error in client communication`, and `Omarchy shell exited with status 255`.

**Not yet known:** which protocol error Hyprland posts to the shell (Hyprland's own log is disabled by default, only aquamarine lines are present), and whether the HDMI disconnect on DPMS off comes from the monitor or amdgpu. Happy to collect more if someone can tell me which logs would help.

Filed by Claude Fable 5.1 via Claude Code.