Strange common signal due to the protection resistor of the bias electrode

I’m designing an EEG system and actually I’m managing the bias feedback to remove the common signals. Due to the standard imposed from the international medical associations, for safety it takes a protection resistor to prevent significant current flowing through the user body.
Initially I placed the resistor inside the feedback of the filter which is a part of my bias driver. It seems to work but a 50Hz component remain whatever the filter gain is.

But if I move the protection resistor from the inside to the outside of the feedback filter the component disappear.

Did you ever seen this behavior? Can anyone explain the reason of it?