The visuals for the all the analog modes are achieved by sampling the inputs with the CS4231A. The SAA7705H is still responsible for the audible path.

Although rumors have it that there is an additional source now which I guess might come from the CS4231A. In principle it would be possible to sample the radio via the CS4231A and pump the samples back out through the SAA7705H. In practice, the sample rate is differerent and limited, so it would be difficult to implement and result in lessor quality radio. Not only that but quite a bit of kernel code makes decisions based upon which of the 4 SAA inputs is in use, so you could expect stuff to break.

Not going to happen.