Can you show us an approximate schematics on what is planed ?
Something like this,
I like the idea of having the switches in series with the resistors for simplicity - we get a single-ended output voltage referenced to AGND.
The disadvantage is that the circuits are sensitive to any non-linearity in the switch over the voltage range.
But - almost all the voltage drop is across the resistors and not the switch. So we're really only concerned with very small voltage differences - as you already noted.
Additionally, having all the switches with one side tied to gnd means we get a common S (eg for Jfets) which makes it easy to drive G with respect to GND and not some floating/changing value.
I am considering dual n-mosfets with sources tied, or relays for lower value resistors (eg. 1-100ohm).
higher value resistors I am considering jfets or analog switches.
The additional bank relay r1 is to isolate high-leakage devices like mosfets - where leakage is on the order of 1uA.
Jfets and common analog switches have leakage on the order of 1nA.
Some analog switches are available with much lower leakage - eg. pA range - and they may be useful for higher valued resistors (eg > 1M) although their RDS(on) is > 1k.
I am considering using an addition bank-like relay on the higher valued resistor bank. A limitation of signal relays that I've found - such as the Omron GS6 is a minimum current of 10uA - so that would preclude their use for shunts over 1M.