Thanks for explanation and links... my lesson for the day!
You're welcome.
Here's my take on it, and the way to visualise how things might pan out. The graphic below represents a dynamic range covering several stops, from -10 below the 0 mark on the camera's meter, right up to +6 stops above the 0 point. Typically you will find that cameras will start clipping data at exposures brighter than +3 above the meter. However, between +3 and +4, if you shoot raw you can typically recover quite a bit of data in that zone. Beyond that the data is pure white and is destroyed permanently. If your exposure looks too bright, but contains all the data you require, you can darkenn the image to make it look correct, and at the same time you will be pushing the noisier data deeper down into the invisible blacks, where it will not show up.
Below the 0 mark, the deeper the tones become, the less signal (light) there is and the more intrusive noise becomes, both that occurring naturally in the light itself, plus that which is part of the camera's own electronics. To the far left the noise becomes very bad, and at the extreme left there is no signal at all - everything is pure black and cannot be recovered.
The light blue bars below the main block represents how a scene containing 10 stops of dynamic range might fall within the camera's capability to record those tones, depending upon how brightly you choose to expose. As you increase the exposure you start to push the highlights into the "recovery" zone, but go too far and you end up clipping the brightest parts, which become a pure, featureless white. However, by increasing the exposure you are lifting the shadow details higher above the noise threshold.
Conversely, as you reduce the exposure more and more, you are wasting the dynamic range offered by the camera at the highlight end (the pink areas), while plunging your shadow data deeper and deeper into the noise, ultimately losing it forever as unrecoverable black. If you try to brighten the image in post you will end up brightening the noise in the "noisey zone" along with the image data. Anything that was recorded as pure black can not be restored.
It's a crude and simplistic graphic, perhaps not even terribly accurate (although not miles form the truth), but hopefully it helps illustrate the point.
The greater the dynamic range of a scene (think sunny day or any harsh lighting with deep shadows and brilliant highlights), the more important it is that you ETTR in order to maximise the DR offered by the camera. If you have a scene with very low dynamic range, such as a foggy day, when everything is just sort of greyish, then it is not so important to squeeze the last drop of DR from the camera. An exposure centred around the 0 mark should be just fine. However, underexposing is never the best option to pick if you have alternatives available.
By the way, in my basketball shot, posted earlier, you may notice that I pulled down the exposure in Lightroom by 0.7 stops (2/3 stop if you prefer). From this, and looking at the histogram after the edit, you might infer, correctly, that I was ETTRing for all I was worth. At f/2.8 my aperture was wide open, and there was no way I was going to try to shoot basketball at shutter speeds below 1/250. My choice on ISO was to shoot at 3200, and never touch the right hand side of the histogram, or bump it up to 6400 and squeeze my capture into the raw recovery zone, knowing I could pull the exposure back down in post. In so doing, the darker areas of the image, with noise already being of concern, could at least be reduced and dimished, rather than being amplified by brightening them in post.
That's loads of light! 
