In a switched capacitor system, the common mode usually gets determined by the prior stage common mode voltage.
And that, in a fully differential system is determined by the CMFB reference of that amplifiers ouput stage.
If you have a system where the input of an amplifier is "open switch floating," give some serious thought to swapping the two inputs to a common mode refernce voltage instead.
As a general rule, I never let inputs to an amplifier "float" because that can get unpredicable. (Especially true with high leakage 90nm - 45nm CMOS) It may simulate nicely, but expect a rude awakinging when you get it in the lab.
A lot of the academic textbooks don't mention that the inputs should be swapped to center scale DC reference points when unused. It is kind of like all the literature on bandgaps does not bother to deal properly with startup circuits.
In a similar fashion a lot of books illustrate ground referenced designs. That is suicide from a noise perspective, which is why the world does everything fully differential if at all possible.
THe devil is always in the details.