Here's a really interesting idea floated by Andrea Cau (with help from Mickey Mortimer, Ville Sinkkonen, Rutger Jansma and Zach Miller) over at his Theropoda blog.
Everyone is making a big deal about Balaur bondoc, the apparently double-sickle clawed dromaeosaur. The double sickle idea comes from the very strange nature of the foot (close-up image in the last post, linked above).
Above: Fossils of Balaur bondoc reconstructed as a standard dromaeosaur. Credit: PNAS/AP.
The first sickle claw is present on the first toe. Normally, in non-avian theropods, the first toe is small and placed high on the foot, like the dew claw of a dog. Basically, it's vestigial. However, in lineages where the first toe evolves some kind of use, the entire metatarsal supporting the toe tends to descend, placing the first toe at the same level as the others. This is seen in two lineages of theropod: birds, where the first toe is used for perching, and therizinosaurs, where it's been re-adapted for walking. Animals that use four toes in walking (rather than having four toes but only using three) are called functionally tetradactyl.
So, did Balaur not use its enlarged first and second toes to kill prey as in other dromaeosaurs, but simply as extra walking digits, as in therizinosaurs? Balaur is certainly built like a therizinosaur, with short, stocky hind limbs and very weird hips. The hip bones of Balaur, as you can see in the image above, are extremely swept back. Some of this may be due to crushing, but probably not all. Normally, such a hip arrangement is seen only in herbivorous dinosaurs, which need to clear out space in the torso for their expansive, plant-fermenting guts.
Tim Williams on the DML has also pointed out that the atrophied hands with reduced third finger and fused wrist elements are also indicators of decreased predatory ability. The only other dinosaurs I can think of off hand that have such reduced third fingers are avialans and the herbivorous Caudipteryx (and, of course, alvarezsaurids and tyrannosaurids, but for different reasons). Avialans and Caudipteryx are the only ones that retain fairly normal hand proportions while shrinking the third finger.
Much has been made of the fact Balaur lived on isolated Romanian islands, and that the "island rule" probably had a lot to do with its weird anatomy. Could similar effects have led to this line of dromaeosaurs becoming fully herbivorous? Weirder things have happened to island dinosaur lineages. Just look at the kiwi.
The parallels between Balaur and therizinosaurs are hard to ignore, but obviously it would be helpful to have some skull material before definitely drawing any conclusions about its diet. In the mean time, check out the cool reconstruction Andrea included in his post for what a therizinosaur-mimicking dodo-raptor may have looked like.