Unhidden by 0.4.15-dev-2069-gfd8080b, Google Earth 7.1.8 regressed AGAIN by the same work

To reproduce

• Use VBox 5.1.x to 6.0.x, or VBox4.3.12 (not 6.1 so far)
• spend 1024MB at least to the VM (I did), better even 1500MB
• Enable 3D support in graphics settings
• Install ROS
• After graphical setup, boot to safe mode and install VBox Additions with "Direct3D" support enabled.
• Back in normal mode switch desktop resolution to 1024x768 32bpp
• install Mesa driver 17.01 from rapps
• Install Google Earth 7 using offline installer , e.g. 7.1.8. from here
• (if you have just 1024MB RAM then another reboot is a good idea here to not run out of mem later)
• then run Google Earth, either in OpenGL mode or DX mode, should not matter, it used to work both

Expected result:
It should run properly in OpenGL and DX mode and you should see the swirling globe, like it did in:
reactos 0.4.13 release
0.4.14-RC-36-gb30846a
reactos-bootcd-0.4.15-dev-2068-gea26bef-x86-gcc-lin-dbg.iso_part3_DX_OK.log
reactos-bootcd-0.4.15-dev-2068-gea26bef-x86-gcc-lin-dbg.iso_part3_DX_OK.webm
reactos-bootcd-0.4.15-dev-2068-gea26bef-x86-gcc-lin-dbg.iso_part4_OPENGL_OK.log
reactos-bootcd-0.4.15-dev-2068-gea26bef-x86-gcc-lin-dbg.iso_part4_OPENGL_OK.webm

Observed result
no swirling globe, content just stays blank, also no "tip of the day"
the window remains blank and checking the menu you will see that the application is not yet logged on to the Google server. Clicking the menu item for logon does nothing.
it does work neither in OpenGL mode, nor in DX mode
reactos-bootcd-0.4.15-dev-2069-gfd8080b-x86-gcc-lin-dbg_AFFECTED_NOOPENGL_NODX.log
reactos-bootcd-0.4.15-dev-2069-gfd8080b-x86-gcc-lin-dbg_AFFECTED_NOOPENGL_NODX.webm

For the record, we already broke this app once already by commit 0.4.14-dev-1239-g7481bda and then fixed it again with 0.4.14-RC-36-gb30846a (for the release) and with 0.4.15-dev-730-gabbe656 for master. That work was covered in ticket CORE-17106. It was matthiasbasler who reported the regression back then. And today it is me reporting it, because after a lot of testing, that we did here for this topic, we decided, that it makes more sense to have 2 bugs fixed instead of just 1.
And also it is much easier to fix CORE-16757 later, because we have 100% open-source then (only ros), whereas fixing Google Earth later is much harder, because we do not have insight in the source-code then!