but with higher bit intermediated computations you do get the best out of that sow's ear I won't argue with that! I think if it's clear that you will go through a heavy online postproduction workflow (Heavy grading, Animation VFx etc.) and disk space is no problem, then you should definitely convert the footage to higher bit-depth aswell as with better quantization (most people use 5dtoRGB for that, converting to, say ProRes as you probably know, DNxHD is a lossy codec and since you stated that disk size is a factor to consider, I still think 120 will suffice. You can't make a silk purse of a sow's ear is the old expression. Which is why the Adobe apps pretty much assume doing all computations in 32-bit floating if they can. But with most mid/lower price video cams we are essentially still shooting 8-bit jpegs. using 10-bit processing on an 8-bit jpeg doesn't compare at all with say the 14-bit RAW file of a D3, quite true.

The resultant images, back in 8-bit, had smoother transitions that were VERY obvious when you had enlargements to make. It was easy to test that for simple changes, there was no difference, but heavy duty work in Photoshop definitely benefited from taking the time to change to 10-bit. When digital first came in, the only thing we could shoot was 8-bit jpeg. I've been a stills portrait pro for 38+ years. The vastly raised precision of the intermediate computations allows for a final result that is at least somewhat better than you might have had otherwise.

even with an 8-bit image, there are a lot of times where the processing of it may be complex that using 10-bit for the processing 'stream' is quite useful.

Xmedia recode lossless 10bit to 8bit