The ol' "only affects users who leave the lens cap on" bit may make you feel better about your 5D3 but even a cursory investigation will demonstrate that this is simply not true. There are plenty of examples where this design/manufacturing flaw (take your pick) makes a difference. Here's one:
http://www.cameratown.com/reviews/canon5dMKIII/light_leak.cfm
Now obviously this isn't an issue that's going to affect most users most of the time. If it were, it would have never made it past the testing period w/o being corrected. But it can be a problem and should be fixed. Smug comments implying that people actually taking photographs won't be affected is wrong.
You do realize that the behavior that's shown in that link is how ALL TTL SLRs behave. Canon, Nikon, Sony, Pentax, it doesn't matter. Nor does the model. When you remove your eye from the view finder, you allow more light to enter the viewfinder which is where the meter is, as a result the meter reading changes. This has been the case going back to well as long as cameras have had metering mounted in the viewfinder and is, in part, why SLRs have built-in viewfinder shutters or covers to black out the viewfinder.
Edit to add:
This is exactly the problem with "cursory investigations" when it comes to cameras, or any other complex system, when you don't control all of the variables you have no idea what is actually going on, and any conclusion you draw from the test is just as likely to be incorrect as it is to be correct. And just because you don't think the environment is such that there is sufficient light entering the viewfinder to affect the metering doesn't mean that there isn't.
