## What happened

`quickshell` (Omarchy's shell) segfaulted at 08:43:45 while the machine was resuming from suspend and outputs were being torn down/re-announced. Quickshell's own crash handler caught it, wrote a report, and relaunched (~0.5s, no visible impact beyond a tray-icon reconnect).

Symbolizing the core with `gdb` + Arch's debuginfod (full debug info resolved — this is not a guess from offsets) gives:

```
#5  atomic_load_p (a=0x0, ...) at include/jemalloc/internal/atomic.h:82
#6  rtree_child_node_tryread (elm=0x0, ...) at src/rtree.c:110
#7  je_rtree_leaf_elm_lookup_hard (... key=1 ...) at src/rtree.c:237
...
#11 ifree (tsd=..., ptr=0x1, tcache=..., slow_path=false) at src/jemalloc.c:1944
#12 je_free_default (ptr=0x1) at src/jemalloc.c:2060
#14 g_free (mem=<optimized out>) at ../glib/glib/gmem.c:208
#15 output_handle_name (data=..., wl_output=..., name=0x... "eDP-1")
        at ../gtk/gdk/wayland/gdkscreen-wayland.c:1813
#19 wl_closure_invoke (...) at ../wayland-1.26.0/src/connection.c:1243
#20 dispatch_event (...) at ../wayland-1.26.0/src/wayland-client.c:1732
...
#24 _gdk_wayland_display_queue_events (...) at ../gtk/gdk/wayland/gdkeventsource.c:201
#25 gdk_display_get_event (...) at ../gtk/gdk/gdkdisplay.c:442
```

jemalloc is asked to free pointer `0x1` — not a real heap address — inside GTK3's legacy Wayland output tracker (`gdkscreen-wayland.c`), handling a `wl_output.name` event reporting `"eDP-1"` (the internal panel). That code only runs inside quickshell's process because Qt is told to load GTK3 as its platform theme: `hl.env("QT_QPA_PLATFORMTHEME", "gtk3")` in `/usr/share/omarchy/default/hypr/envs.lua:14`. GTK3's own Wayland output bookkeeping appears to hold a stale/dangling pointer that gets freed when re-triggered by the burst of monitor add/remove churn during suspend/resume (journal shows repeated "Got removal for monitor ... which was not previously tracked" warnings in the same window, on `FALLBACK` and then `eDP-1`).

## What's Omarchy's and what isn't

The invalid free itself is upstream, in GTK3's Wayland backend, not in this repo's code. But GTK3's Wayland output tracking only executes inside quickshell because Omarchy sets `QT_QPA_PLATFORMTHEME=gtk3` for theme tracking — without that, quickshell wouldn't load GTK3 at all, and this code path wouldn't run.

## Expected

Resuming from suspend with external monitors attached and the internal panel present shouldn't segfault the shell, even transiently.

## Steps to reproduce

Not confirmed as reliably reproducible — happened once on this machine. Pattern: laptop with an external-monitor setup (internal `eDP-1` panel present but not in use — DP-3/DP-4 external monitors active), suspend, then resume. Within the output re-enumeration that follows, the crash lands on the `eDP-1` output's `wl_output.name` event.

## Search performed

Checked `omacom/omarchy` for existing reports (`quickshell SIGSEGV`, `gtk3`, `wl_output`, `jemalloc rtree`, `gdkscreen-wayland`). The large #7380/#9843 family covers a different, unrelated failure: a clean `exit(255)` from a Qt Wayland protocol error when the last real output disappears under a session lock — no coredump, no GTK3/jemalloc involved. This is a distinct bug: a genuine SIGSEGV with a GTK3-specific invalid-free stack.

## Environment

- Omarchy: `4.0.2-1`
- Hardware: Razer Blade Stealth 13 (Late 2020), hybrid Intel Iris Xe (i915) + NVIDIA GTX 1650 Ti (nvidia 610.57.04)
- Monitors: internal `eDP-1` (off), external `DP-3`/`DP-4` active (2560x1440 each)
- Hyprland `0.56.2`, quickshell `0.3.1`, Qt `6.11.2`
- Kernel `7.1.9-arch1-2`
- Full diagnostics: `/tmp/omarchy-debug.log`

Filed by Claude Sonnet 5 via Claude Code.