The height sensor is an inductive type sensor. Their measurements are affected by changes in temperature as well as magnetic fields, both from the DC current flowing through the bed heater traces and the permanent magnets.
The working theory found on many forums is the most accurate readings would be obtained with calibration done without any bed heater current, at a constant temperature throughout the entire calibration, and also with no measurements taken directly above the bed’s permanent magnets.
The strength of these effects is debated widely, from imperceptible up to deal breakers making some (non snapmaker) machines unusable.
Early on in the snapmaker firmware history a quick z calibration check was preformed automatically before printing, after the bed was heated. This had to be scrapped as it resulted in the toolhead crashing. It was presumed at the time this was due to the presence of current in the bed heater and/or the change in bed temperature compared to the previous cold calibration. After further testing it seems now the engineers are saying the effect of DC current induced magnetism is smaller than the thermal expansion of the aluminum platform, so a hot calibration even with the bed heater current on would be more accurate for printing than a cold bed.
Details on the exact sensor used can be found here Anyone have the proximity sensor part # - #50 by dstarke
Edit: there is also debate about the influence of magnets on the accuracy of the sensors. This post was intended to be a summary of previous discussion / literature review of other forums across 3D printing where this had been discussed. I have not verified the accuracy of the information.
I do agree that an air core eddy-current based induction sensor should be insensitive to magnetic fields.