I've wondered that myself, but it may simply be the way the Quark operates. I made
a thread a while back showing the differences on how the MM behaves vs the quark when the motor is forced to a stall condition. Basically, the MM simple increases the throttle until the motor finally moves (which is bad IMO and the root cause of a couple of them burning up) vs the Quark being MUCH more gentle about it. I know this is mostly firmware programming, but maybe the engineers at S&T decided to decrease the slew rate a little for control reasons? And if you decrease the slew rate, there is more time where there is an actual voltage drop AND current draw creating more power loss. I really don't think it's a matter of raw FET specs. But, even in your example, 10W is actually quite a lot of power to be dissipated on a small heatsink.
And, I've not found a reason to get crazy with heatsinking - just enough to pull the heat away from the case and dissipate it to the ambient air, which should have enough flow in a moving vehicle to do so effectively. I don't like heat - at all - so when I say the ESC isn't hot, it isn't. To me, 130*F is too much.