1. HDR is not to blaim, it's the (IMO over-the-top) processing, tonemapping, that causes it.
2. Some prefer postprocessing that's over the top. These images are often lacking in other aspects (e.g. composition) as well, and the effect is an attempt to make something usable out of nothing.
3. Well executed (HDR) tonemapping is almost invisible to the viewer (like with the old master painters), it just looks natural.
I've done alot of thinking on this subject recently. What has me fired up AGAIN, is the misapplication of the technique.
Before I go on, to address your point 1, "Tone mapping" is using your computers software to change one set of numeric values that represent color and luminosity in your image to another set of values representing different colors and luminosity. Any process that changes these values is technically tone mapping (so you can see it covers a wide range of functions). There are many ways to do this; curves, hdr, selective color, desaturation, split toning, etc, etc. All of these are tone mapping techniques. However, what most people mean by "tone mapping" is strong/excessive areas of local contrast with an overall reduction in global contrast. The "HDR Look".
When I stated photography, we used the wet dark room. So for me, what we could do in the wet darkroom is the "gold standard" when it comes to processing. That includes dodging, burning, double (or more) exposures, use of filters, push/pull processing, etc. There is quite a bit of latitude there, and modern software just makes it easier. My point is, I don't have any problems with folks post processing their images...
So as I was saying, I was thinking last week about what HDR does for us. During the course of that mental exercise, I thought of the HDR process in a new way. Depending on your particular camera and sensor, modern DSLR's give us about 10-11 stops of dynamic range. The human eye can resolve between 17-20 stops of dynamic range, depending on conditions (where the light source(s) is relative to the eye). Let's say we take a three exposure, -2, 0, +2 set in anticipation of doing some hdr work. When we do we are adding about 4 steps to the dynamic range to the photo. So less assume 11 stops of dynamic range from our camera, and 4 more from processing and that's 15 stops, and is getting pretty close to what the human eye can naturally resolve. If we use a wider spread of exposures to build our HDR, then it's even closer.
Where my beef with HDR users lies is that many of the scenes we see on flickr or elsewhere do not contain 15 stops of dynamic range. If they don't, there is no technical reason to do an hdr shot of a car, tractor, building etc, etc if the shot is taken in broad daylight with good lighting. Now if you are doing HDR for artistic reasons then you might have a valid reason for the HDR treatment. If it's simply to "expand the dynamic range of the photo", it's pointless as you camera can already capture the entire dynamic range present in the scene. It's inefficient, a waste of processing time and storage space. (granted the processing time can be minimal, and storage has never been cheaper).
It's also important to note, that while HDR images expand the dynamic range of the scene, don't really "expand" the dynamic range of the final image that we see. Most HRD software up-rezs the composite images to a 32 bit composite. If my calculations are correct that's 4,294,967,296 colors. This is a true expansion of the original Dynamic range. However, since our monitors and printers are generally 24 bit sRGB displays or 16 million colors, they can't display the 32 bit images, do the software DOWN-rezs them to 24 bit 16 million color images. This down rez effects both luminosity and color depth and compresses both. Both the uprez and downrez are "tone mapping".
To me, photo's with lots of dynamic range have a significant amount of contrast as they are reproducing a wide range of light to darks. Most HDR software REDUCES this contrast by balancing the histogram and spreading the light and dark areas across the entire frame. This is done by making highlight areas darker and shadow areas brighter. This is done on a pixel by pixel basis by the HDR software analyzing the individual pixels themselves. Basically an inverse s-curve is applied to every pixel in the image, which results in a global reduction in contrast. This allows the image to be displayed by media with limited (24 bit, 16 million color) dynamic range. The result is an image with balanced amounts of light and dark areas. The result is an evenly exposed photo, but evenly exposed photos don't result in powerful high contrast images.
So, I'd say that if you don't have 14+ stops of dynamic range in a scene, HDR might not be your most efficient method of reproduction. So let's say a scene has more than 14 stops of dynamic range and I still want to do a single exposure capture? What options do I have. Well, by using a ND grad and by shooting raw, I can balance the original exposure in camera. Then by using Adobe Camera RAW/Photoshop I can selectively lighten the dark areas and darken highlight areas by hand (dodging/burning or local adjustment brushes in ACR), any other areas that might be too light or too dark. One can easily push/pull 1 stop out of a raw file and you can usually get 2. So conservatively, that puts me at 16 stops, which as I said before, is very close to what the human eye can naturally resolve.
HDR is a tool, and like any tool you have to know how it's best applied, and when it might be best to use another tool...