We have been suffering about this problem quite a long time, maybe since the beginning of starting to use nRF51822 roughly two years ago. The thing is that we cannot trust that the nRF51 consumes only a few microampers after it is powered up.
In a typical scenario when measuring current consumption of power-down state we can achieve a few microampers as expected (below 10 uA or so). However, quite often the board "decides" to become a monster power-eater and we get several milliamps consumptions, from 3 mA to 5 mA. Re-powering the device does not typically help. The most reliable way to get back into the microampers is to short-circuit the board after removing battery, and re-connect the battery then. However, even that trick does not help always.
This is a very critical and fatal problem for our business, since we cannot trust our battery-powered device runs a year as we have calculated according to microamper measurements. Instead, if the moody little device happens to start in "milliampers mode", the life-time of battery is only a few days, or less!
Any ideas what could cause this problem? Is it related to nRF51822 debug-mode as in some early chip versions? Or could it have something to do with reset pin? Of course, our external circuits may be causing the problem, but what could be the mechanism and why it does not happen always? And actually, there's not very much more than the nRF51 chip itself. What kind of measurements would help us to find out the root cause?