Oh, guess I should have said. It's a simultaneous sampling ADC- not muxed inputs. Muxed inputs wouldn't be such a big problem since that ADC has an output for the mux and an input for the ADC so I could just use one fancy signal conditioning assembly.
EDIT: Oh, I think I might have a solution...
Since a differential ADC has a dynamic range twice as large as it's absolute input range, I can use a voltage reference half as large with a single-ended input to make use of all ADC codes. I still connecting the output voltage of the sensor to the positive ADC input, but instead of connecting the sensor ground for the negative ADC input, I use the bias voltage of the sensor or my desired common mode voltage.
So a gyroscope with a 0-5V output and a 2.5V bias for 0deg/sec would have the ADC measure the differential voltage between it's output and 2.5V bias. If a reference voltage larger than 2.5V was used then you start losing dynamic range since the gyro output never strays more than 2.5V away from the 2.5V bias. So instead of using a 5V reference (which would be ideal for a sensor 5V differential output), I would use a 2.5V reference to measure single ended signals. In actuality I'll probably use 2.048V reference and give up a bit of dynamic range. Since the ADC has an analog 5V power source it can still tolerate any input signals whose absolute values are between 0-5V even when it's reference voltage is less due to the magic of differential inputs.
Of course, with a 2.048V reference, if I measured the differential voltage between a 0-5V signal and ground I would get clipping past the reference voltage, and lose half the negative portion of the dynamic range. But most analog sensor signals are bi-directional so they have some bias that the output voltage can be measured with respect to.
Differential inputs are cool...if not requiring lots of parts.