I appreciate the idea of the soft-max scaling with 7.4 but mostly because I realize I need something else… the “just like that but different” category 
I normally avoid setting the Y-axis maximum so the dashboard is effective for a wider range of systems. Even for things that have a natural maximum (like 100%) selecting or excluding some of the series is more effective when auto-scaling is used.
The problem that I have with auto-scaling is that a single big spike in my data sets the sale and renders the panel useless for the rest of the time. Quite often the exact height of the spike isn’t that important to show (depending on the exact time a brief spike occurs, the sample average can be a factor off anyway).
Wouldn’t it be possible to have an option to scale the axis such that for example 95% of the samples fit on the chart. When looking at an hour in 1-min samples, there could be 3 off-scale, maybe even less when the scale is rounded to nice numbers afterwards. You could still hover over that spike to see the actual value when it’s off-scale.
