Tim, would there be an advantage to replacing the video op amps with discrete parts?
I presume you meant “Ted” instead of “Tim”? 
Tho for audio bandwidths you can implement a discrete opamp with some features that better most integrated opamps, video opamps have some more challenges:
Video opamps operate at much higher bandwidths. The ones the DS use need to have a useful bandwidth of, say, 400MHz - they have to be linear (have the same gain and delay) from DC to over 400MHz with as good a signal to noise ratio as possible over that range.
Audio opamps (integrated or discrete) can achieve a signal to noise ratio > 120dB over the audio band but doing this at higher frequencies is extremely difficult.
Video opamps (at least the ones I use) are inherently balanced (differential input and differential output) Tho you can double up the circuitry in a discrete implementation getting them to perform exactly differentially (identically in levels and in delay) at higher bandwidths is next to impossible.
I also use some weird features of video opamps - in principle balanced opamps don’t care about an common input DC offset between the inputs: since the output is an amplified version of the difference between the inputs any common mode noise (including a DC level) will be canceled out. Video opamps also usually let you specify the common mode DC offset of the output - kind of a strange feature that requires more than the two obvious simple differential amplifiers when implementing a video opamp.
Also the DS uses four copies of the digital switch and associated low pass filter to convert it into audio in parallel for each input channel. So I’d need at least 16 identical discrete opamps (and possibly another 8 if I use the settable output DC offset.) And in an ultimate DS there’d probably be at least a factor of two more.
Yes, sorry Ted.
That does make sense.