I found that if i use a UV LED flashlight against a BME280 temperature increase dramatically and humidity decrease the same way, instead pressure seems not affected.
The spectrum of LED UV flashlight is this one, taken with an Ocean Optics USB2000+ UV spectrophotometer
Is the sensor so sensible to UV light?
Yes the UV light goes directly inside but I discovered that the UV flashlight warm all surfaces, also my skin. I don't know if it is an effect of UV or the flashlight has also an IR emission. White LED flash doesn't have this effect but I don't feel warm if I use it on my skin.
Anyway the graph shows that it is not a photoelectric effect since the warming and the cooling are not immediate.
We should avoid any direct light into sensor hole.
It will cause internal MEMS behavior goes to wrong like increased temperature slowly.
So please cover sensor and test again.