Georgia and I went down to the X-end to quickly enhance the sensitivity of the EFM.  We were marginally successful, but unfortunately discovered more problems.  

First and foremost was an large peak at 80 kHz in the Y arm (Picture 1, the main curve is the Y arm spectrum, the reference curve is the X arm spectrum).  There was also a 900 kHz peak we read with an Agilent.
On May 31st, at around 18:05:00 UTC, the Y arm spectrum becomes much worse, with a large peaks everywhere and high RMS.  This corresponds with in-chamber closeout work where the EFM may have been knocked around and some circuitry could have come loose/been shorted.  

We took an SR560 and plugged the EFM X arm differential output into it with a gain of 100, then routed the output into the ADC.  We did this because our CDS EFM noise was clearly limited by the ADC noise floor of 4 µV/rtHz above 100 Hz.  

Plot 2 shows the ADC output before and after the gain has been applied to X arm, as well as the messy Y arm spectrum.
Plot 3 shows the EFM CDS channel calibrated into volts/meter before (in blue) and after (in red).  
We win a little bit of sensitivity to electric fields by avoiding the ADC noise, but also amplify the EFM sensor noise by too much. 
 
Plot 4 shows that our EFM X arm CDS channel (purple) is still not at the same level of sensitivity as the SR785 (blue).  We will have to go back down to the X-end and be more clever about this.