Recover screen sharing that stopped working in every app at once
Read hyprwm/xdg-desktop-portal-hyprland#437 in full (open). It supports runSync with no timeout, the single-threaded wedge, the Failed to close session implementation: Timeout was reached log, the FALLBACK output and restore-token mechanism, and the custom_picker_binary workaround. Confirmed locally that xdg-desktop-portal-hyprland.service exists as a user unit, xdph is 1.4.1-2, and Omarchy's /usr/share/omarchy/config/hypr/xdph.conf sets custom_picker_binary = hyprland-preview-share-picker (package hyprland-preview-share-picker 0.2.1-1), and pgrep -f share-picker matches both picker names. One overstatement: the record presents a locked session OR no powered output as independent triggers. The issue observed only the combination, a monitor power-off that is a DisplayPort hot-unplug (not DPMS) with the session locked, and its author says explicitly that the locked-with-monitor-on case was not tested. Cause and symptom narrowed to what was observed. Fix and verify kept. Not exercised: no wedge was reproduced here.
Symptom
Screen sharing suddenly fails everywhere: Share in Meet, Discord or OBS does nothing and no picker appears. journalctl --user -u xdg-desktop-portal repeats Failed to close session implementation: Timeout was reached. It started after the monitor was powered off and the session was locked while something (often a remote-desktop client reconnecting) asked to share the screen.
Cause
xdg-desktop-portal-hyprland runs the share picker synchronously, with no timeout, and is single-threaded. If a ScreenCast request needs the picker at a moment nobody can answer it, the picker process never exits and the whole portal stops serving every other client until it is restarted. The case observed in hyprwm/xdg-desktop-portal-hyprland#437 (xdph 1.4.1, Hyprland 0.56.2, open) is a monitor powered off, which on DisplayPort is a hot-unplug so Hyprland replaces the output with FALLBACK, combined with a locked session. A client's restore token still names the old output, so xdph discards it and prompts, and the picker cannot be seen or used. The report's author did not test a locked session with the monitor still on. The report used hyprland-share-picker. Omarchy 4 points custom_picker_binary at hyprland-preview-share-picker, which xdph launches the same way, so it is likely affected the same way, but that was not exercised.
Fix
Look for a picker that has outlived its request, kill it, and restart the portal backend:
pgrep -af share-picker
pkill -f share-picker
systemctl --user restart xdg-desktop-portal-hyprland
Then start the share again from the app.
To stop it recurring, close or pause remote-desktop and other screencast clients (RustDesk, OBS with a portal source) before locking the screen or switching the monitor off. The issue also describes a custom_picker_binary wrapper that answers without a GUI when no output is connected, set in ~/.config/hypr/xdph.conf, which is only worth building for an unattended machine.
Verify
pgrep -af share-picker prints nothing before you start a share, and clicking Share now opens the picker.