Turning on triple buffering pretty much eliminates any performance hit from vsync. With it on, it isn't a question of 100-30 vs 75-30, because at 75 you've got power to spare, so if the demand goes up the card can still render at 75.
The reason double-buffering speed drops with vsync on is because of the way rendering works. With double buffering, there are two buffers; what is on the screen, and what is being drawn. With double buffering, you waste rendering time waiting for the next sync cycle to occur. Once one occurs, you swap the backbuffer into the frontbuffer. But at that point, you do NOT have the next frame ready. You've got something displayed on screen in the frontbuffer, but an empty backbuffer that needs to be drawn.
With triple buffering, there is always a frame waiting to be drawn, and the drawing happens in a third buffer. So while you're waiting to swap out the next frame, you're already working on the one after that.
The downside of triple buffering is that it does delay rendering by one frame worth. At a constant 75hz, triple buffering delays video output by about 13ms. Normally it is impossible to notice this delay since there are already other delays at work (so-called mouse lag, for example, is always present, even if it is imperceptibly low). Another delay factor doesn't usually matter. However, there are many circumstances when it IS noticeable.
It is a tradeoff, really. There are three options. the first option is to leave vsync off. The benefit of this is that you get the 75FPS (effectively drawn to screen), and no loag, but the downside is the tearing. The second option is vsync on with double buffering. The upside of this is that it eliminates tearing, and does not introduce any additional lag (not counting that introduced by a lower framerate). The downside is that performance suffers because the card is wasting time waiting for swaps to happen. The third option is vsync on with double buffering on too. This eliminates any performance penalty from vsync, but adds the latency.
One could also easily argue that the 13ms of latency by delaying a frame is counterbalanced by the additional times between frames with vsync and double buffering. Dropping from 75 to 50 FPS would mean about 7ms longer between frames drawn. So the apparent delay only increases by about 6ms by enabling triple buffering, once you take that into account.
As for the issue about 75hz... Well, from the biological perspective, my take is that the human eye IS limited to 30 or 40 hz, except that ignores exposures. Think of a blurry moving object in a camera. Even though you only have one "frame", the object moving is present in multiple locations in that frame, hence the motion blur. While the human eye might only work at 30 or 40 hz, you can perceive smoother (more natural) motion at higher framerates because the higher the framerate is that more reference points there are for exposure. Theoretically, simulating motionblur would not be required on a computer displaying images (without motionblur) at a high enough refresh rate. The human eye, presented with a large number of images per second, would create the motionblur as it does in nature. I suspect this is why people can tell the difference between 60FPS and 200FPS. There is an increase in fluidity.
Another example of exposure would be to think of a picture taken with a shutter time of 100ms. That would be one tenth of a second. You might be tempted to say that any event that happens that takes less than 100ms would be missed. But instead, the film is exposed to that object for a fraction of the exposure time. The image captures a ghost-like translucent image of the temporary object. The eye works the same way. Even if you only see a new image every 30th of a second, objects present for less time still can make an impression.
On the issue of if it really matters, I'm going to say no, not really. 75FPS is high enough to create a SUFFICIENTLY fluid movement. It isn't perfect, but it is "good enough". Advances in realtime motionblur (see what Valve has done with DoD) may further improve this with existing framerates. The eye may not be able to tell the difference between natural motionblur and sufficiently good artificial motionblur. And by this I mean good motionblur. Some motionblur effects simply overlay the last few frames onto the screen. This isn't real motionblur, and it looks like crap.
I don't recall what method Valve used for DoD, but the trailer video is quite informative. If you look at individual screenshots, they look insanely blurred. You think, this is horrible, why would anybody want to play like this? Then you see the footage itself, and all of a sudden it looks very natural. The blur, instead of decreasing quality or sharpness, actually increases the apparent fluidity of motion, making it EASIER to track objects.
Here is a download link to just the motionblur part of the demo footage:
http://www.fileshack.com/file.x?fid=8105
As I said, pause it and it looks blurry. Play it and it looks very natural.