One of the main disadvantages is that a normal Darrieus rotor has a negative Cq-lambda curve for low values of lambda. So the generator must also be able to work as a motor to start the rotor.
Have you looked at the helical variants, such as the one in the link from the original post? As I understand it, they are self-starting.
As I understand it, the problem with starting a Darrieus is that the apparent wind has to be coming from a narrow angle at the the front of a blade for the flow to attach to the blade. When operating, the flow then enters the interaction at a small angle to, and leaves about parallel to, the chord of the blade. Even though a little velocity is lost from drag, the net change in the wind direction imposed by the blade makes the velocity component along the chord higher when leaving than arriving, and thus the reaction force tanget to the circumference is forward.
When the blades are spinning at high TSR, the apparent wind is within the narrow attachment angle all the way around the cycle. The blades get power except for the small part of the cycle when they're moving almost directly up/down wind, and the gain from redirecting the wind is smaller than the drag. But at lower speeds the wind is only attached for part of the cycle, so it's only powered part way around and loses output.
When stopped the wind hits the blades from some random set of angles and they're mostly not particularly well powered due to lack of wind attachment (or if they DO happen to be facing the right way to pull some power, they move to another angle and lose it again.) The rotor MIGHT happen to start up, especially if there's a gust (so they're very dangerous unless braked). But they usually don't. You have to have a mechanism to apply some other twist, to power them through the angles where they fight against spinning rather than cooperate, to get them to run.
But with the helical design (1/3 turn or a multiple per blade), when stopped, some small region of some blade is always at an angle to achieve attachment and forward thrust, (ditto another portion where the wind is backward to the airfoil but it still works a little bit). On the averge you have equal amounts of ALL angles of attack, and the lift from the little patch that's got attachment is enough to overcome the average of the wrong-way lifts and drags from all the other angles. You get a little forward force, which starts the rotor turning, which increases the "from the front" component of the apparent wind, which increases the size of the patch that's powered, and thus (if you don't load it down until it's moving adequately) the mill spools up to operating speed in a positive-feedback loop.
Now there are lots of OTHER issues with a Darrieus. A big one is vibration, from the constant change in amplitude and direction of forces as the mill goes through a cycle. The helical design mitigates the vibration of the mount substantially because, on the average, the blades are equally at all angles all the time. But the force on any patch of the blade still cycles with the rotation, and the center of effort moves along the axis. So you still get some flexing and some vibration applied to the supporting structure. (Even if you used two sections with opposite twist you'd still get a bending moment as the centers of effort move in opposite directions.)
So material fatigue is more of an issue than with a HAWT, where the net forces are usually pretty constant at any given set of operating conditions and change gradually. With the (necessarily) spindly structures of the Darrieus variants, the components aren't well supported against these constant cyclic forces (unlike the less efficient Savonious variants.) Thus the simplicity of not needing tracking is, IMHO, trumped by the likelyhood of "unscheduled disassembly" and other vibration issues. Though the Darrieus is closer to the efficiency of a HAWT, making a robust-enough design to survive the weather is tougher than just getting tracking and furling right on a HAWT, and being able to trust it is not in the cards for me.
