Another issue I've run into with this board was the microcontroller's power supply.
Since the voltage range of a LiFePO4 cell is compatible with the voltage range of the MSP430 microcontroller, I powered the MSP430 directly from the bottom cell. This would ensure minimal power consumption, which is very important to preserve the battery's shelf life and to make sure the battery doesn't get discharged further once undervoltage protection kicks in. Since LiFePO4 cells also have very low internal resistance, I figured they would act like a big cap to shunt any noise that would be generated by the booster.
Wrong! Well, maybe the cell works as a *big* cap in that it has similarly bad high frequency behavior. :) It certainly did not filter out switching noise and the MSP430 did not like it (oddly, the micro behaved kind of unpredictable and timers seemed to be running *faster*--maybe they were being clocked by the noise?). It seems the cells have very low internal resistance at DC, but the impedance increases with frequency.
Hacking in a ferrite bead and extra filter capacitor on the back of the board solved the power supply noise problem.
Another problem with the micro's power supply is that during charge, the extra 0.2V regulation voltage across the current sense resistor pushes the voltage at the top of the bottom cell up to 3.8V toward the end of the charge cycle. This is still well within the absolute maximum rating (4.1V) of the microcontroller, but outside the recommended operating conditions (3.6V). I'll probably revisit the power topology in the second rev.