If I may jump in on this, I suggested the value of 20 based on a few trials with the first THERM test I made (Fairyfly user feedback), looking for a balance between high-ish mesh density and short-ish simulation time.
The issue was that a lower setting was messing with the final mesh color display when there was a high temperature jump between two points: my example showed the mesh jumping from black to green while skipping the purples and blues due to a lack of intermediate points. If I remember correctly the previous mesh parameter was set to 5 or 10. I went up to 50 on the same test, but the simulation was taking a few seconds too long for an unnoticeable change in mesh quality.
I also noticed that smaller shapes get a much higher density of points than larger ones, so probably a lower setting might be ok with details containing a lot of small shapes, like your example above. At the end of the day what matters is a perceived acceptable simulation time, I guess.