We are currently working on an energy harvesting solution that powers the nRF52832. The energy harvester is providing energy from indoor solar and at the lowest light level (200 lux) the solar cell outputs close to 20uA at 2.5V. The energy harvester has a built in boost converter that can charge the attached two 220uF capacitors to 3.65 volts, and will enable the output of the capacitors to power the nRF52832 once they charge up to 3.3 volts. This leaves around ~400 uC of charge that can be used by the NRF at boot before discharging down to below the operating voltage of 2.2 volts.
The 400uC of charge has been more than enough to boot most of our modified applications. However, when merging the secure_bootloader and another application the boot process uses more charge than running the two independent of each other, as seen by the PPK output:
Only the bootloader on the device:

Only the application on the device:

Merged bootloader and application using mergehex and nrfutil:

These measurements were taken by using the PPK and the PCA10040 DK. The PPK would be connected, started, and then the power output would be enabled. The highlighted regions represent the first fractions of a second when the power output was enabled.
My guess is that there's some kind of loop that is drawing a large amount of charge. I'm not exactly sure what would require a wait cycle that lasts 1/8th of a second while transitioning over to the app.
Any guidance or help to get rid of this high power state would be wonderful, and thanks in advance.