### System details

- Laptop: Lenovo ThinkPad X1 Carbon Gen 13 (Intel Arrow Lake-U, iGPU [8086:7d41], driver i915; the log says "Found meteorlake/u")
- Omarchy 4.0.4-1, kernel 7.2.5-3-omarchy, Hyprland 0.56.2
- Encrypted root (stock Plymouth + encrypt hook, stock early `thunderbolt` module)
- Dock: Lenovo USB-C Mini Dock (VIA Labs hub). Same behaviour with several other docks from different brands, all known good.
- Monitor: Samsung Odyssey G91SD, 3840x1080@60 (EDID is read correctly, 256 bytes)
- Second machine with the identical symptom: Dell XPS 13 (2024)

### What's wrong?

With a USB-C dock and external monitor attached, the monitor worked once. From the second reboot onward it never comes back in a usable state, with or without the dock attached at boot. The internal panel always works.

What I see on every boot now: the monitor gets a picture for a few seconds during session start, then the output disappears. Nothing I tried brings it back: reboot with the dock attached, reboot without it and plug it in later, replugging.

### What the logs show (failing state)

The external monitor starts as `HDMI-A-1`, is modeset successfully, then the connector drops and the same monitor re-enumerates as `DP-1`, which can't be modeset any more.

Hyprland log (aquamarine), in order:

```
drm: Connector HDMI-A-1 connected
drm: Description Samsung Electric Company Odyssey G91SD (HDMI-A-1)
drm: Modesetting HDMI-A-1 with [email protected]
drm: Modesetting HDMI-A-1 with [email protected]
drm: Connector HDMI-A-1 disconnected
ERR drm: Cannot commit a disconnected output
drm: HDMI-A-1 is not connected, clearing stale crtc 269
drm: Connector DP-1 connected
drm: Description Samsung Electric Company Odyssey G91SD (DP-1)
drm: Modesetting DP-1 with [email protected]
atomic drm request: failed to commit: Invalid argument, flags: ATOMIC_ALLOW_MODESET ATOMIC_TEST_ONLY
WARN drm: atomic commit failed with max_bpc set, retrying without max_bpc
```

That TEST_ONLY failure repeats for every mode (314 failed commits in this session, down to 640x480).

Resulting state:

```
$ hyprctl monitors all
Monitor eDP-1 (ID 0): [email protected]
Monitor DP-1 (ID 1):  [email protected]   (availableModes lists [email protected], [email protected], ...)
                      disabled: false

/sys/class/drm/card1-DP-1/status:     connected
/sys/class/drm/card1-DP-1/enabled:    disabled
/sys/class/drm/card1-DP-1/dpms:       Off
/sys/class/drm/card1-DP-1/edid:       256 bytes
/sys/class/drm/card1-HDMI-A-1/status: disconnected
/sys/class/drm/card1-DP-2..4/status:  disconnected
```

hyprmoncfgd keeps retrying and never succeeds:

```
apply failed: context deadline exceeded: DP-1 mode mismatch: wanted 3840x1080, got 0x0
DP-1 (Samsung Electric Company Odyssey G91SD) kept coming back without a mode, and it is already at its gentlest settings
```

Kernel side: no `ucsi_acpi`, `typec`, `i915` DP/AUX or link-training errors in the journal for this boot (only the usual "Selective fetch area calculation failed in pipe A" at i915 init). Type-C ports bind normally (`typec port0: bound usb3-port3 / usb2-port1 / usb4_port1`). So the kernel logs nothing, but the DP connector refuses every atomic modeset. This is different from #11908, where `ucsi_acpi` logs `GET_CABLE_PROPERTY failed`.

### Steps to reproduce

1. Boot Omarchy with the USB-C dock and monitor attached; the monitor works for that boot.
2. Reboot (dock attached or not).
3. From the second boot on, the monitor shows a picture for a few seconds, then disappears; `hyprctl monitors all` shows the output at `0x0`.

### Related issues (none matches exactly)

- #13904 and #13020: external output stuck at 0x0 after suspend/resume or after the monitor daemon restart. Same `0x0` + `Invalid argument` signature, but those trigger on suspend/restart, not reboot.
- #13073: NVIDIA DP outputs stuck at 0x0 (CRTC/connector mismatch in the kernel's atomic check).
- #11908: Meteor Lake, dock monitor lost after suspend, `ucsi_acpi GET_CABLE_PROPERTY failed`; reboot fixes it there, not here.
- #13211 / #11019: dock display not re-detected after replug (kernel 7.2.x).

What seems different here: it happens across reboots (not suspend/replug), it reproduces on two laptops from different vendors and with several docks, and a reboot does not clear it.

I can test kernel parameters, other kernels or a stripped-down Hyprland monitor config and report back.