Looks like a good idea. How can I verify that these devices do not include things that might alter the signal accuracy, like filters? Ideally a straight transformer with nothing added would seem right. I do want a wide bandwidth, at least to the Nyquist limit on my current sampling rate (48 kHz sample rate, so 24 kHz bandwidth). That should not be saturation issue (even below 60Hz as long as it is not a strong signal).
Another idea I thought of since posting was to build some kind of self contained sensor that could isolate the hazards better, and have it do PWM or PCM of an LED to pass on the voltage readings. I understand some LEDs can be modulated on/off well past 1 MHz. I'd then need some kind of device to pick up this light and pass it into the computer. Maybe a transformer box like shown could be the host for that LED (removing the LV cord).