LTC2943 stops responding when discharging current

Hello,

I have a problem with the LTC2943 fuel gauge.  The description of the design is as follows.  I have a 5 cell NiMH pack that I am charging up from a 7.9V boost supply.  The battery pack charges up to a max of 7.5V.  Becuase I'm charging fairly slow the cell voltage peak is not pronounced and I usually only see around 7.3V or so.  The charging works fine.  I have an MCU communicating with the LTC part and while charging everything looks great.  The I2C communication works fine, all the registers are as expected and the charge current the LTC part is giving me (accumulated charge called ACR register) looks good. 

My issue is discharging.  If I discharge over 70mA the LTC2943 seems to shut down.  It stops responding to I2C messages and I get a under voltage lockout UVLO bit set in the status register when it comes back to life.  The only way I can bring the part back to life is to reduce or turn off the load.  When my load is off or very load everything is back to working like normal.  The datasheet says the UVLO condition occurs when the voltage on the SENSE+ pin drops below 3.5V.  When I measure this on a scope I see no drop outs and my battery sits consistent around 7V.  I've tried multiple battery packs. 

1. Are there any other ways to get a UVLO condition? 

2. Is it expected to see the I2C digital side stop responding?  It is very vague in the datasheet.

3. The datasheet says "when the SENS+ pin drops below 3.5V without reaching the POR level".  Is there a way this can happen with 7V on the SENS+ pin?