Is it possible that the speed of the shutter action is a variable that depends on mfg batch, phone battery voltage, or something else?
With decent scope maybe it will be possible make own DIY using only this part cut from whole PCB - sensor itself and put there MPU we like
This above it not of course thermal image of sensor part of PCB, but just for fun applied this quite nice
iron256 LUT which is very similar or maybe even based on FLIR's one mentioned many posts above
While using even basic image processing methods I was able get decent output based on this USB raw sensor data provided in this thread earlier by @marshallh, the most interesting thing is simply hack or learn its protocol between MPU and sensor, and maybe even make... own DIY shutter and forget about this builtin in-simply remove it
In my Linux app added some adjustable gamma and can controll how smooth output image is, so with this
iron256 LUT it looks much better than those very noisy looking
Seek Thermal outputs and it is time to make OpenCV version of this software, while probably I was able to find quite good way to fill those hexagon
black dots and other probably bad pixels, so it could be nice make this code more efficieint and accelerated by Nvidia CUDO technology supported in OpenCV.
Sample PNG RGB output after 4x resizing oryginal sensor 208x156 size to output size below to compare with those images taken from Android devices