A lot of verbiage, much of it technically accurate, but in many places missing the point for practical picture taking. I'll leave the rest of it without further comment.
On this point, however, I want to say something further.
Sensors exhibit very substantial sample variation. Like many other electronic components, sensors are binned by quality. Grade 0 sensors used in scientific and technical imaging have essentially no defects (e.g., hot and cold pixels) beyond a stringent threshold. Rather boring scientific monochrome cameras with grade 0 sensors (say, a 1.4 megapixel front-illuminated microlensed interline transfer Sony) are generally in excess of $8000, and often sell for $15000 or more.
Sensors of that bin quality will not be found in mass-produced pro/consumer cameras. These devices will therefore exhibit subsstantial sample-to-sample variation.
A RAW file contains no data that describes the pixel-level defect profile of a specific sensor, and cannot compensate for this variation. An in-camera JPEG engine can easily store and utilise such an individual profile. In fact, in scientific imaging, for critical applications, we always profile the individual sensor, even though it's generally grade 0.
And that is why a good in-camera JPEG engine can, in principle, in some applications, trump the RAW developer on your computer.
