When the vehicle is in motion shock from the road is transmitted up
to the roof rack. When a bicycle is clamped by the fork, into a
fork mount, the fork and headset are held rigidly in place. As the
vehicle goes down the road the headset will be constantly absorbing
small shocks. This would not be a problem if the headset were also
turning, because the lubricant inside would be distributed by the
rolling of the bearings. But the headset is not turning and
therein lies the problem.
From:
http://draco.acs.uci.edu/rbfaq/FAQ/128.html
"The damage occurs when these small motions occur when there are no
steering motions to replenish lubricant while the bearing balls
fret in place. Fretting breaks down the lubricant film on which
the balls normally roll and without which they weld to the races
and tear out tiny particles." This was written by Jobst Brandt.
However, it is the small vibrations, smaller than the Hetzian contact
patch between ball and race that cause fretting damage. "Rolling
Bearing Analysis" by Tedric Harris (SKF) explains this in fine detail:
http://tinyurl.com/y38had
On upright mounts the wheels are absorbing the shock from the road
but on fork mounts the shock is being absorbed by the headset
though it is unknown how much extra wear occurs because of
fretting.
It's not the shock, but small vibrations that damage the bearings.
You can pond on the tires with a hammer all you want and not cause a
fretting dimple. This has been discussed here at great length.
For CF, the forks are so susceptible to damage you don't want to be
holding the bike by them.
There is lots written about this ....