Same problem here (since a few monthes, but never reflashed since then).
I add a few more details hoping to involve more people.
What happens is that the lower the current used to charge, the higher the percentage of charge reached BEFORE the red led
blinks and the charging process stops: i.e., charging with wall charger leads to 80% (battery charge), with usb port 70%.
After a few experiments I am now able to fully charge my mobile. I had to install i2c_tools and a slightly modified (by
me) version of joerg charging script. The modified script stop BME, loads the appropriate values on the bq 24150 charging
chip and keeps it away from being watchdog-stopped by resetting it's internal clock every 15 seconds (but doesn't read
any of the bq27200 values).
I can guess that when the problem occurs the bq27200 fails somehow, and any attempt to read it's values (battery temperature, charge percentage, etc...) leads to a fault. In fact, with the battery fully charged if the script gets stopped and BME gets started the phone will struggle, suddenly becoming unresponsive, and I guess that the problem is due to BME trying to read bq27200 values but not being able to.
Fortunately it looks like bq24150 keeps working properly, though ensuring a proper and hopefully safe charge.
I have to say that I had big problems with my phone, since it loose the phone signal probably after an heavy rain. It was then repaired by a third part technical center (since Nokia refused to repair it). The battery issue strted a few weeks after the phone was repaired, so it might be related (but it might be not...).