Does the light leak when the mirror is up or just when it is down?
I tested with the mirror up shining a 3 D-Cell maglight at the LCD at point blank range, and get no appreciable difference between two 1s exposures at ISO 25,600. So I'm confident in saying that no, the light leak, at least on my 5D-3, doesn't seem to reach the sensor.
For that matter, with a lens on the camera even in a dark room (metering <1/4, f/4, ISO 6400 / or <EV0) I can't get the backlight to alter the exposure in any appreciable way. (I.e if it's teetering on the edge between 1/4 and 1/5, it does so regardless of the backlight state.) For that matter, I see more fluctuation in the meter due to subtle camera movements changing the framing and the orientation/positioning of my eye to the viewfinder than there is from turning the LCD backlight on/off in these tests. Likewise, shooting into the dark from under a light and shading the LCD is making no discernible changes to the meter for me.
I should note my 5D-3 fails the test as given in the article. Likewise, the shine a light on the LCD test, not only does my 5D-3 fail, but so does my 1D-3 and I've never had an exposure/meter issue with that in the 2+ years I've had and used that body. Moreover, they seem to meter near identically for the same scene, lighting and framing, at least within my ability to control for errors.