This is a good point and I suppose will come with time and practice.
I started with these not very nice hihat samples. They are all the same velocity, but different openness. And wrote a matlab script to cross fade them. They all have the same cross fade period and in this picture it's 50ms. So the fade in is 50ms, there's 50ms of audio, then 50ms of fade out. The fade in is not applied to the first hihat sample and the fade out is not applied to the last hihat sample. Just to be clear.
Hopefully it's obvious that the timescales are not the same for the left and right half of that image.
The original and the combined cross fades I tried shorter periods, but it wasn't as noticeable.
If you can, I'd love to test it.