Hangs with "started gnome display manager.." after CUDA 9.2 install with TB3 eGPU

After a CUDA 9.2 install on Ubuntu 17.10 and 18.04 on a NUC6i7KYK with a Thunderbolt 3 HP Omen eGPU the system hangs with

started gnome display manager and dealing with any system changes

With the HP Omen Thunderbolt 3 eGPU off, I originally got past the login, but then the mouse pointer (only visible item on purple) froze.
Solved this in switching from “Ubuntu” (meaning Wayland) to “Ubuntu on Xorg” in the cogwheel menu under the password input.

Any chance to get this running? What problem does Gnome have?

On Windows 10 1803 the eGPU works nicely with CUDA 9.0 which is currently needed for TensorFlow 1.8 and the driver loads dynamically, but I want Linux with nvidia-docker and there doesn’t seem to be any intention to port this (soon, though Docker is Linux on Windoze as well).

Edit1: Additional tidbit(s):
#1 Just read and confirmed they changed the default in 18.04 back to Xorg (see “interesting” naming in the cogwheel menu ) and Wayland is not to be blamed for everything, but the other weirdo move from Unity back to Gnome.
#2 IMPORTANTLY after the nth restart (variant), having it got to run with eGPU off it wants to report the error:
Failed to upgrade/install trying to overwrite ‘/lib/udev/rules.d/71-nvidia.rules’ which is also in package nvidia-kernel-common-396 396.24.02-0ubuntu0-gpu18.04.1
As it occurs to me I had the system running with CUDA 9.0, but wanted the more recent driver for nvidia-docker.
Edit1 end

Thanks for any help!
G.

This thread and the solution is very much related to
https://devtalk.nvidia.com/default/topic/1036640/cuda-setup-and-installation/cuda-9-0-install-on-ubuntu-18-04-quot-unmet-dependencies-quot-/post/5269814/#5269814

Abstract:
Many pitfalls with eGPUs and then some with CUDA installation for DL frameworks.

Exotic stumbling block:
/etc/X11/xorg.conf needs an AllowExternalGpus

Section "Device"
    ...
    Option "AllowExternalGpus" "true"
EndSection

For details see the other thread.
G.