This part was tricky to do. What I ended up with was having it zoom so that something like 5 samples either side of the split point are visible (I forget the exact number). This makes the zoom depth depend on the surrounding context, which worked better than a fixed zoom level. But it still isn’t perfect, because sometimes the samples will go from being clumped to suddenly spread out, which can make the zoom depth a bit hectic when nudging back and forth near those spatial changes.
Yeah I was hoping you’d mention this! That’s something I want to add too. It would also b cool if it could automatically identify those points in the path where the samples cluster, even when the sample types aren’t stationary. Then it could intelligently offer easy to tap visit extraction points.
That wouldn’t actually be all that difficult to make work, and would be significantly less fuss to use than the current two-splits approach.