I can disregard CPU because it actually doesn't even max out on the approach. Thing is this issue is old, very old and it's noted on the P3D forums (Despite having no attention from the devs), it can happen in a number of airports, mostly payware with a big aircraft such as the 747 or C130.
What caught my attention was that the ORBX manuals say that it doesn't add a new mesh BUT that it REQUIRES a mesh resolution of 5m although i don't understand why.
I found this on the ORBX manuals:
"5m terrain mesh resolution is the slider
setting that all Orbx regions and airports are
designed for. That being said, using either
10m or 2m are also unlikely to lead to any
major display issues unless you look very
closely. Since the slider setting governs the
horizontal spacing of mesh grid points this
value primarily affects the steepness of
slopes adjacent to flattened areas at airports
or lakes and rivers. If objects are placed on or
next to such slopes they can shift a bit."
This issue has been reported on a number of payware airports and i have encountered it several times. But one that particularly catches my attention is ELLX, when landing on RWY 06L, the payware developer put something inside a hole next to the runway, and it causes that problem to happen. I believe the way P3D handles collisions is directly affected by this.
http://www.prepar3d.com/forum/viewtopic.php?f=6314&t=131079&sid=bcf15675b8a8c9097e788c64a9361f5a This thread has a number of tests and people who have done a lot of research in this. Maybe this could help FSDT understand what could be colliding with the 747 and perhaps offer a fix, assuming that's even possible.
Now i have tested the KORD 10L approach many times, i have a video of the EXACT moment that collision happens and i believe it's the objects right next to the runway that are doing that, i can upload the video if it helps.