I'm mostly going to address one point raised in Matt Grum's answer and (especially) the comments on that answer.
On average, a MF camera/back does give more control over depth of field than a smaller format camera can. With the larger format, you don't need as fast of a lens to get just as shallow of depth of field. A "full frame" MF camera has a sensor roughly 1.7 times the size of a full-frame 35mm-sized sensor. At the same focal length you need to be just over half as far away to get the same framing, which reduces the depth of field considerably.
@Jędrek Kostecki commented that you often don't want shallow DoF. This is only sort of true at best. For a lot of product shots, what the designer often (usually?) wants is a depth of field that's almost impossible to achieve just by adjusting the aperture. With normal adjustments, the depth of field follows a pattern something on the general order of a sine wave -- at some distance, you get maximum sharpness, and then becomes progressively less sharp as you get further away from that ideal distance. With a large aperture, the sharpness falls over very quickly, and with a smaller one it falls off more gradually.
What the designer often wants wants is more like a square wave: essentially perfect sharpness for the full depth occupied by the product, combined with an extremely fast reduction in sharpness for all other distances.
There are a few ways to do that. One is to manipulate the setting, so everything else in the picture is either a lot closer, or a lot farther away than the subject. The problem with this is that it frequently leads to rather unnatural-looking perspective. To combat that, you can shoot from a quite a ways away with a long lens.
Just for example, I shot one like this with a 600mm lens (so old the focal length was marked as "24 inches") on a 4x5. It worked, but it was a pain to set up -- in that case, we had the camera way up in one corner of the studio (had to climb a ladder to get to it), the background at the diagonally opposite corner something like 60' away, and the subject object on a stand roughly halfway between. It was impressive looking though (I was looking at the GG, so I didn't get to see it, but I was told that the client walked in just as I was pulling the dark cloth over my head, incidentally "unveiling" that giant old polished brass lens barrel -- from the rather indelicate description I was given afterwards, he almost...um...wet himself on the spot).
A much easier way to do it is to set up all the objects at roughly normal distances, use a wide enough aperture to get the fast falloff of focus you want, and use focus stacking to get the amount in focus that you want. For small format cameras, focus stacking is mostly the domain of a few lunatic fringe macro specialists and such. With MF cameras, I've seen focus stacking used even for landscapes.
Of course, I would probably be remiss if I didn't also at least mention the fact that quite a few MF cameras also have T/S lenses available, or (in the case of a tech camera) have at least limited movements built into the camera (e.g., rise/fall is fairly common). This, again, lets you control DoF in ways that aperture alone just can't duplicate.
Edit: (mostly in response to Matt's comment):
This is not even close to purely theoretical. For example, let's consider the Hasselblad 100mm f/2.2. At the risk of being accused of cheating, I'll figure things up for its closest focus (90 cm), using a CoC of .03 mm (fairly standard).
In this case, I get a total DoF of .97 cm.
Looking at the Nikon 105 f/2, and increasing the distance to maintain the same framing (approximately, anyway -- it can't be identical, since the sensors aren't the same proportion), I get a distance of ~160 cm. At that distance, (and using the same CoC) I get a DoF of 2.6 cm -- well over double the DoF with the Hasselblad.
If I go to the Nikon 85/1.4 instead, I have to move closer, to about 130 cm away, and I get a DoF of 1.86 cm -- down to just a tad under double the DoF of the Hasselblad.
If I go to the Canon 85 f/1.2 instead I maintain the same distance, but the larger aperture decreases the DoF to 1.56 cm -- only about half again more than the Hasselblad.
Assuming I could find a Canon 50 f/1.0, I'd move in a bit closer yet (76.5 cm away), and my DoF drops to 1.31 cm.
About the only possibility that's left would be to use the Leica 50 f/0.95 instead. That would drop the DoF to something like 1.25 cm, but it's still definitely more than the Hasselblad.
I do also feel obliged to point out that to use the Leica Noctilux 50 f/0.95, you're starting to get into MF-style pricing as well. The lens itself goes for $10,495, and to use it, you need a Leica M-series camera -- I believe the sole (digital) choice is the M9, which goes for $6,995. Assuming I've typed the numbers into my calculator correctly, that works out to $17,490 for the pair.
On the Hasselblad side, an H4D-31 (with an 80/2.8 lens) goes for $13,995, and the Hasselblad HC 2.2/100 goes for $3,255, giving a total of $17,250.
Bottom line: the guy on DPReview was wrong. An MF will produce shallower DoF, and even coming close with a 35mm sized-sensor isn't exactly cheap either -- and the Canon option is the one that requires a lens that's really hard to find.
Edit (mostly as a more complete reply to @coneslayer's comment): I've held the CoC constant because what's interesting here is the characteristics of the cameras themselves, not characteristics I might choose to impose on the results I get from those cameras.
If I change the size of the circle of confusion from one camera to another, then the result you get from the computation is almost entirely one of my own choosing. It's no longer based on the characteristics of the cameras themselves, but simply of my own judgement about the final result.
In other words, it becomes a matter of my creating depth of field by fiat -- deciding that if X amount of unsharpness was produced by one camera, that it still qualifies as "sharp", but if it came from the other camera, it's "not sharp".
When you start to do that, you can get whatever result you decide you want. The result you get no longer has much of anything to do with the cameras themselves at all -- it's just my preconceived notion, fed into a formula and turned into a number to disguise my idea as if it was an objective fact.
Now, I don't mean to attack the basic notion of varying the CoC to suit the kind of print you're going to produce. If you decide you want to be able to print a particular picture at a particular size, and have it meet some specific criterion for sharpness, that's perfectly fine.
While it's certainly true that the camera/lens have some effect on the decisions you might make in such a case, it's also true that most of the driving factors are the decisions you make, not the camera or lens. If you want to know something about the camera and/or lens, you need to factor your own judgements out of the equation -- and in the case of computing DoF, the only real way to do that is to hold the CoC constant.