@mojo-chan - We are using a dedicated fuel gauge IC, though I have not noticed any problems. I've attached graphs of the current and voltage during a charging cycle. There is a little bit of noise, but this data is from the fuel gauge. There is a bigger problem that we have with it. If the battery is discharged to the point of the protection PCB cutting off, the fuel gauge's memory resets (it's volatile) and loses the record of charge accumulation. Strangely it doesn't reset to zero. I plan to compensate for this by implementing a Voltage -> Charge look-up table with interpolation and current correction. Then the % will no longer require charge accumulation. The reason I have stayed with a coulomb counter is that I really like having the current and voltage sensing functionality. Now I'm asking myself why I'm using a "fuel gauge" at all... We are currently undergoing a major redesign, so I'll look into other alternatives.
@ivan - It seems to have no problems when attached to the USB. The draw on the battery decreases dramatically, but is still present on account of my USB voltage being about 4.85V. We are thinking of using a buck regulator attached to the battery voltage to provide the 3.3V. Though, at this rate, we may have 3 switching regulators in the final design. We are planning to implement the unlocks before sending out any energyShield, though the receive date will most likely get pushed back a little. At present, the energyShield is the exact footprint of the Arduino and the DC jacks line up. However, we are considering different for factors for the energyShield. Possibly making it standard shield length, possibly sitting a little lower. This is just discussion at the moment. Which would you prefer?