To test if the check if mtime_orig != mtime_new: is indeed the problem, I made an operator (simplified version of the original addon_enable(): developer.blender.org/diffusion/B/browse/master/release/scripts/modules/addon_utils.py$327) with that removed:

The smoke display was rewritten in 2.8 and does not suffer from these types of artifacts anymore.

Archiving Blender Internal report, we will no longer focus on fixing bugs in it now that it was removed in 2.8.

There's not really any corruption, just the same versioning issue as far as I can tell.

dustBin's displacement in //assets/dustBin.blend on the rim object (called Cylinder) is making problems.
I assume this is still not using the 'new' displacement node in that file (it will be there when you open that file, but that is probably because cycles adds that automatically when seeing an old file)
And when linking, this automatic insertion fails, see T54504.

that "portal" seems to be the dustbin object...
checking whats wrong with that...

deactivating compositing deactivate the hang/crash, but strange "portal" still there anyway after saving:

@Jacques Lucke (JacquesLucke) if I don't save the file (use the original directly), it just hangs forever, but no crash anymore. I still have to kill as it will take 1 core of CPU forever(waited 5minutes at 25%) just after rendering. Actually, the crash was "better" as I didn't had to kill the process myself. It seems it's due to the compositing phase?
However, the file corruption when saving still happen with this strange portal being added in the left corner of the room.

I cannot reproduce the crash.
Does the crash happen in the newest Blender 2.8 build, even when Render Samples is set to 1?

Was re-reported as T56591, closing.

found a pretty recent buildbot which worked: 78a8d36 from 12 June 2018

There's no bug here:

• To make numpad 1/3/.. switch to ortho mode, enable Auto Perspective under User Preferences > Interface.
• To have the result of the boolean visible when you move away the sphere, you have to apply the modifier first.

looked some more into this, looks like chrome ran into a similar issue coming from math.h, all i can say is 'yikes'

It has to do with cycles motion blur, 2.79b seems to render fine.
Disabeling motion blur or setting it to Center on Frame seems to render the boolean OK.

Proper way to deal with this would be to move every CPU kernel into a namespace. That would ensure linker never re-uses "wrong" microarchitecture implementation.

Confirming it's caused by above mentioned commit. @Campbell Barton (campbellbarton): feel free to assign to me and I'll try my luck :)

afaict, this is also fixed by rBfb3528d08870, thx @Campbell Barton (campbellbarton)!

afaict, this is also fixed by rBfb3528d08870, thx @Campbell Barton (campbellbarton) !

Can confirm this, issue in undo. This can probably be merged into T54578, T54553?

Similar to T54553 (py API doesnt set TextUndoBuf), similar thing as D3139 will fix the crash here, too:

In T54464#491074, @LazyDodo (LazyDodo) wrote:

seems like you have forgotten to attach the file.

seems like you have forgotten to attach the file.

Discussed with other developers about ways to process here.

I'm aware of this, it's an intentional trade-off we are making. With Russian roulette you get more noise in simple cases, but it helps in more complex cases like transparent hair. The two are difficult to distinguish for the renderer and it's generally not helpful to do so in real world scenes where you have a mix of these.

Fabien, you need to recompile blender - best build blender-v2.79a-release, this will be the official stable release soon.

Thanks for your help
How can I apply the fix? Can I simply copy and paste fixed files?