The ADC part of the question isn't really different than any other type of device. Transformers are really the only choice if you need to handle multiple volts of common mode offset. You pay for that though, I just checked at Parts Express and a Jensen input transformer is about $75. You can get a TI INA1651 for $1.
The THAT 1200-series InGenius® high-common mode rejection balanced line receiver ICs overcome a serious limitation of conventional balanced input stages: poor common mode rejection in real-world applications.
That graph you posted in the technical documents section made it sound like perhaps you thought you could just connect a line input directly to an ADC, but that will not be acceptable performance in real world use. Generally you would want to put an instrumentation amp in front of an ADC chip to handle common mode noise and convert between traditional line levels and what the ADC device will expect. Keep in mind that evaluation boards from device vendors are for evaluating the device, so they put as little as possible between input/output pins and connectors. Those designs are not something you would want to use as-is in a final product, but the vendors don't want to have to address questions about why the eval board doesn't exactly match the specs in the device datasheet that of course refer to the raw device, not device plus conditioning circuitry.
The particular chips mentioned by the OP include a specific front-end that provides the necessary signal conditioning. So it's a basic low-level balanced input that is designed to receive directly and electret of MEMS mic. It can also be used as a "line input" with limited headroom, which indeed requires additional conditioning to allow it to interface with conventional audio stuff. It would be some kind of attenuation and EMI/RFI protection. So you're absolutely right: "The ADC part of the question isn't really different than any other type of device."
In normal operating conditions, nothing objectively justifies the addition of a xfmr.