Intro
Following suggestion of DAB I have decided to verify my sensor's readings by installing a second one (and then a third one...) and got some - unexpected to me at first - results.
Second sensor and a surprise
Another sensor was installed on the same shelf as a first one but in opposite back corner. When the first readings appeared, they were somewhat unexpected:
After initial cooling-down period, second sensor (yellow line) started to provide data of very different envelope than a first one. Does it means that a first sensor is malfunctioning? This curve is more closely resembling data from posters promoting correct placement of different types of food inside a fridge (like one below), when uppermost shelf is supposed to have temperature in the range of 8°C:
so there is a possibility of the failure of the first sensor...
Let's check
Sensor shuffling
Below is a schematic plan of the top shelf of my fridge as seen from the top:
This fridge is somewhat unique because it seems to have heat exchanger inside the left wall instead of external radiator at the back (that can be identified by the fact that the left wall gets warm when the compressor is working).
Initially, sensor_1 (that is the name we will use for our Oregon sensor) was placed in the position A, and sensor_2 (reported as Nexus-T by rtl433) in position D.
After some time, another observation was made
At 08:04, sound of compressor start was heard - it means that the compressor starts well before the temperature reaches maximum (09:04), which means our time budget is worse than previously expected - it was previously calculated as time of rising temperature vs. time of lowering temperature. This difference is unlikely to be a timezone problem - we are at GMT+1 currently, so readings incorrectly stamped with GMT would be rather at 07:04 not 09:04.
Then, sensors were moved - sensor_2 was placed in the place "B" at first (in front of sensor_1) - which resulted in temperature drop. Then, another sensor was added and they were arranged in series: sensor_2 in the back left corner (yellow line), before it sensor_3 (position "B", blue line) and finally sensor_1 in front (position "C", green line),
Conclusion
As we can see - temperature distribution inside of cooling chamber is uneven, simply increasing the distance from the wall by about 5cm greatly reduces temperature variation during cooling cycle (yellow vs. blue curve) and raises mean temperature. It also means that the power of appliance may be too low to quickly lower the temperature inside.
And in the meantime we have obtained sensor system with triple redundancy, ready for most demanding use cases