Mr. Borg does bring up an interesting point. Suppose you are shooting a landscape where the dominant color is green. Assume you are shooting at base ISO, and have picked the maximum exposure that doesn't blow out important highlights.
On a whim, you look at raw data from the capture. You happen to notice that while the green pixels (those behind the green filters in your Bayer Pattern sensor) have a much higher exposure than the red and blue pixels. Perhaps if you increased the exposure by one stop the blown pixels would all be green pixels, and none of the red or blue pixels would be blown.
If your goal is to maximize exposure, you could put a filter on the lens that reduces the green light by one stop, and leaves the red and blue light untouched. You could then open up the aperture by one stop, or use a one stop longer shutter speed.
The net result is that your green pixels stay at the same exposure, but you have increased the exposure in the red and blue pixels by one stop.
By increasing the exposure on half your pixels (25% of the pixels are red, 25% blue) by one stop, you have increased total light captured by 1/2 stop.You are now capturing more total light, and therefore will have a less noisy final image. Yes, you would need to compensate for the filter when processing the raw data, but that's a doable task.
So if your goal is to do everything you can to maximize the light captured, then this is a technique you should consider.
On the other hand, at some point the quality of an image is more than good enough. You reach a point where additional increases in quality do not justify the resources needed to get those increases.
All real world captures have noise. If the noise is low enough, then you may be able to measure the noise, but it won't be visible to the unaided human eye. One can make a reasonable case, that once you reach the point where the noise is not visible, there is little benefit to further reductions in noise.
Perhaps you are shooting without a filter, and your images have no visible noise. You could use Mr. Borg's technique to further reduce measurable noise, but that incremental reduction may not justify the time, effort and resources needed. Of course, whether or not that additional reduction is worth it, is going to depend on the situation. In some situations, you may need to squeeze every drop of performance from your camera. In other cases, you may be able to use a simpler workflow, and still get results that meet your needs.