I can't offer anymore to you than the math as proof what you need to do is a fairly heavy edit on both and print both if the difference is not apparent using your PP to your eyes then stay with the 14bit.
Your partially right in your response but I did not claim that 16 bit reduces noise when you say "Sensors are too noisy. We need better sensors or frame averaging to see the benefit" your correct if noise is larger than the size of the quantization steps, then increasing bit depth won't reduce visible noise.
So you say "no-one sees a difference"
Modern sensors are smooth in there tonal response, real world scenes rarely stress all channels equally and most displays and printers heavily mask differences. So yes I agree in many images you won't see banding in 14-bit with good exposures and gentle edits but that doesn't mean the precision isn't there or that it never matters. It means failure mode is rare, not non existent.
Where your argument or comment is technically weak is when you say "All processing happens in the same bit space anyway" I think your missing the fact that you can't invent what wasn't captured and The Higher precision math prevents further damage but it WILL NOT RESTORE LOST TONAL STEPS.
IF THE RAW CAPTURE CLIPS OR COARSE-QUANTIZES DATA: NO INTERNAL PIPELINE FIXES THAT LATER that why is said "16-bit is safer, not that it magically improves images. That being said the harder you push your file(s) the better chance you have with a 16-bit file because 16-bit reduces the risk of quantization artifacts rather than improving noise.
At very high dynamic range levels (around 18-20 stops), the perceptual benefit of an additional stop is small because current displays and prints cannot render it directly and the extra range must be tone-mapped. In practice the value of additional DR at that point is primarily in capture and processing headroom- exposure flexibility, highlight roll-off, and reduced risk of clipping-rather than a clearly visible important to most viewers. if you capture high-contrast landscapes, backlit scenes, seascapes with specular highlights and make large fine art prints as I do it matters. On the other hand it matter much less in portraits, low-contrast scenes and Web-Only output.
In closing my issue with smaller sensors (Sony Pony's and I own them) is not the fact that they can't make large prints it's the loss of resolution when cropping. With digital today I shoot mostly everything wider and then crop to taste, when I used to do a large amount of Macro we had to pull back to gain depth of field so a H6D-100c was my trick pony the file was huge and even though the objects only accounted for 1/3 of the sensor we still had enough mega-pixels to make 40x60" images tack sharp from a given viewing distance.
