### System details

#   Desktop PC, single monitor (Dell S3422DW on HDMI-A-1), Arch,  #   omarchy-dev 4.0.0.r1624.g03b825f, quickshell-git 0.3.0.r20.g28771c7  #   (also reproduced on a Lenovo Yoga C640 laptop, same versions)

### What's wrong?

# "What's wrong?" field: everything below this line

# ---------------------------------------------------------------------------

**Symptom.** After the machine sits idle and locks, coming back gives Hyprland's
*"Oopsie daisy, it looks like you locked your screen but the lockscreen app died"*
screen — and after clearing it, **the bar is gone too**, because `omarchy-shell`
(quickshell) is no longer running at all. Happened **three times on 2026-08-10**
across two machines.

**What the log shows.** These are the final two lines the shell emits before the
process disappears:

```
DEBUG: Not creating lock surface for screen QScreen(0x…, name="") as it is not
       backed by a valid wayland output.
WARN:  The Wayland connection experienced a fatal error: Invalid argument
```

So the lock path encounters a screen with an **empty name / no valid Wayland
output**, correctly declines to create a lock surface for it — and then the Wayland
connection dies, taking the whole shell with it.

**Two consequences, both bad:**

1. The session is locked with no lock client → Hyprland's dead-lock wall, which
   needs a TTY (or another machine over SSH) to clear.
2. The shell is dead → no bar, no widgets, until it is started manually.

**It writes no crash report.** `~/.cache/quickshell/crashes/` stays empty, so nothing
surfaces the failure; the shell appears to have simply evaporated. The only trace is
`journalctl --user | grep omarchy-shell`.

**Not a flaky-hardware artifact.** On the desktop there is exactly **one** connected
output (`HDMI-A-1`), the other connectors report `disconnected`, and there are **zero**
`EDID block 0 is all zeroes` messages in the journal for that day. The invalid screen
appears transiently — most plausibly while the display sleeps or outputs are
reconfigured around the lock.

**Aggravating detail:** Hyprland's recovery text advises `killall -9 hyprlock`, but
**hyprlock is not installed on Quattro** — the lock is the `omarchy.lock` Quickshell
plugin. Anyone following that instruction literally hits a dead end. Recovery is
`hyprctl --instance 0 eval 'hl.clear_crashed_lockscreen()'`, plus starting the shell
again.

**Suggested fix.** Treat "screen not backed by a valid wayland output" as skippable
without tearing down the Wayland connection, and re-arm the lock surface if a valid
output appears later. Failing that, having the shell restart itself rather than exit
would at least keep the bar and the lock alive.

**Workaround in use here.** A 30-second user timer that restarts `quickshell` when it
is not running. It deliberately does **not** clear a crashed lock, since
`hl.clear_crashed_lockscreen()` unlocks the session without a password — so the bar
comes back automatically, but unlocking still requires the user.