I've removed J4 (extra 3.3V header) from the board and connected J7 (VBAT header) to a Keythley 2450 SourceMeter. The board preforms as expected with ~800nA being consumed from VBAT while PMID is also outputting the same voltage as on VBAT but when reaching the battery undervoltage lockout threshold (3V) the current consumed from VBAT jumps to ~3uA. In this mode when 5V is applied briefly to J6 (VIN header) the current drain from VBAT jumps to ~400uA !!! this drain is constant until the voltage on VBAT drops to 2.7V when the current drain again becomes ~3uA.
Is this behavior considered to be working as designed? What can I try to mitigate these issues? As described the behavior would kill the battery our product is using within hours.