There exist several platformers using 3D graphics with 2D gameplay, so no, obviously it isn't an issue. The most recent case that comes to my mind is the new Giana Sisters game.

Also there's the advantage that having a 3D mesh (and its textures) is going to take up a lot less of memory than high resolution sprites (especially as you add more animations), as well as the fact it scales up with the screen resolution without much problem (with sprites you'll end up needing even higher resolution textures).

I would simply consider which you are more experienced in; 3D modelling and animation or 2D animation. Unless you intend this project as a learning experience, in which case doing it the hard way might be more beneficial to you in the end.

It depends on the level of detail you want, too, so if you have 30 different attack animations per character it would probably be easier to deal with a single 3D model and animate it accordingly rather than hand-drawing every frame. That's assuming you know how to model, texture, rig, and animate, however.

Animating 3D objects such as characters is the ideal to be obtained if possible. That being said, some games use lower resolution 2D sprites behind the main characters. With some planning, you can also have distant ones be low resolution sprites working in areas which will never or seldom see the player go.

I would say that a combination of 2D sprites and animated 3D would make a good balance of performance and visual quality, but you must also of course consider your ability to deliver such a strategy.

My 2D game has rather high artwork requirements due to the size of some of our sprites. A few years ago I proposed that to make this process easier, we recruit some 3D artists who can make really basic animations, take frame captures, and then send them to the 2D artists to touchup and make it look like real pixel art. Unfortunately, we couldn't find a single person with skills in making 3D art that was interested in doing this type of work (for free anyway), so we had to abandon the idea and cut down our artwork requirements by a great deal. Basically, we re-use our map sprites in battle and increase their size by 4x. In the long run, I think this was actually the better option to take for our project as having a different set of sprites for maps versus battle was just too much for a free project like ours to handle.

I use Blender to create my sprites. They are not that bad, but not very good at lower resolutions.

The main thing I could say is that in all the time it takes to "re-work" a 3d render to look like detailed pixel art,you may as well have just done the pixel art from the start. But, I'd say if you need references to make the pixel art, then a very simple 3d render could give you that, say if you can't get poses right or something like that.

Also, I believe that the Giana sisters game is actually 3d with 2d gameplay, not rendered using sprites. The OP wants to do a 2d sprite based game, but create the sprites from 3d renders. I could be wrong though,as I don't have the game's source.

Rendering objects to a render target does have it's perks over traditional spritework. For one, as already stated, it can potentially take a lot less memory to store than large or numerous graphics or spritesheets. Animation can be done by simply iterating through the model's frames, and bones can be moved manually to add functionality that would otherwise be extremely difficult to achieve, such as preserving the character's running animation while holding a large object in their hands (animation blending) or looking at something in the world (bone transformation). Single parts of the body can be scaled and twisted, so that you could easily incorporate things like "shrink potions" or "giants" for the player to fight. Because of the perk of easy scalability (no detail is lost when the image is scaled up), you can easily write a small algorithm that makes the sprites look the same on any computer, taking up X percent of the screen's viewport resolution.