Still 32 active bits!
It is the same data stream that goes out to S/PDIF or AES/EBU digital outputs (via CoreAudio) and that I will be listening to on external DAC!
but that can´t be, at one point the float format must be converted to int in order to pass it to a s/d converter.
Where does the float format and extra 16 bits is introduced in the case of that particular app? I am streaming 16 bit fixed without any math involved. It's a CD rip.
Are you saying that in CoreAudio fixed is always converted to float and then back to fixed when going out to S/PDIF?
you will have to decide if you want to test it or not. no test, no evidence. 
I trust my eyes and tools I'm using. I see stuff that's not supposed to be there. That's enough evidence for me.
P.S. If you want, you can have it. Test it for yourself. 10.10+
