On a hybrid Intel(iGPU)+NVIDIA(dGPU) laptop with an external monitor wired to the
  NVIDIA GPU, changing display brightness under Wayland instantly freezes the whole
  session. The NVIDIA-driven external display freezes first, the cursor survives a
  moment on the Intel-driven panel, then total lock. SysRq/VT-switch unresponsive;
  only a hard power-off recovers.

  Regression: 100% stable on Plasma/KWin 6.6.5 for days. Freezes started on the first
  session running 6.7.0 (and 6.7.0-3.1). Downgrading ONLY kwin to 6.6.5 removes the
  freeze entirely (same kernel, same NVIDIA driver) -> regression is in kwin 6.7.0.

  Key evidence it is NOT the NVIDIA driver:
  The kernel fault is byte-for-byte identical across TWO different NVIDIA branches:
    NVRM: Xid (PCI:0000:01:00): 31, name=kwin_wayland, MMU Fault: ENGINE GRAPHICS
    GPC2 GPCCLIENT_PROP_0 faulted @ 0x0_07e40000, FAULT_PTE ACCESS_TYPE_VIRT_READ
  Reproduced on nvidia-open 610.43.02 AND 580.159.04 -> kwin submits invalid GPU
  work to the secondary GPU. Likely same family as bug 521161 (closed as upstream/
  NVIDIA), but the driver-independence + clean 6.6.5->6.7.0 regression point to kwin.

  Hardware:
  - iGPU: Intel TigerLake-H UHD [8086:9a68] (i915) -> eDP-1 (internal panel)
  - dGPU: NVIDIA RTX 3060 Mobile GA106 [10de:2520] -> HDMI-A-1 (external monitor)

  Steps to reproduce:
  1. Optimus laptop, external monitor connected via HDMI (wired to the dGPU).
  2. Wayland session.
  3. Change display brightness (powerdevil backlighthelper) -> instant hard freeze.

  Software:
  - KWin 6.7.0-3.1 (crash) / 6.6.5-4.1 (stable). Plasma 6.7.0. Qt 6.11.1.
  - Kernel 7.0.12. Mesa 26.1.2. CachyOS (Arch-based). Wayland.