I have added a new feature (this PR) that allows you to set a new probing configuration attribute called 'samples_trunc_count'.
What this does is it will remove 'samples_trunc_count' number of the highest and lowest samples when probing and then return an average (mean) of the remaining samples.
This is very useful when using a sample count of greater than 5 samples as it increases the benefit of the additional samples while still being resistant to outliers (unlike regular average alone).
Let me know what you think and if you have any comments 