In the various CPU % usage charts in the shard metrics, it looks like the upper limit of the Y axis isn't fixed, and it changes based on what the maximum value in the range of time being displayed is. For both normalized and regular CPU% values, there's a natural maximum to the value, I think: 100% for the normalized CPU%, and ncores*100% for the non-normalized CPU%.

With a changeable Y limit, it's hard to tell at a glance whether the value is actually really high or not. Oh, the blue line is hovering up near the top of the chart. Is that using a lot of CPU? Dunno; have to glance over at the Y axis labels and read the little scale numbers - oh, no, it isn't; this particular chart only goes up to 20% CPU on the Y axis at the moment. But if I come back later and refresh, maybe the scale will change to 100% and a blue line near the top of the chart does mean a lot of CPU.

Plus, if you view both the plain and max versions of the CPU metric on the same page - e.g. the Normalized System CPU and Max Normalized CPU, those two charts may not end up with the same Y axis scale. So you can't easily compare them visually. E.g. here's an screenshot attached where Normalized CPU was scaled to about 25% on the Y axis, and Max Normalized CPU was scaled to about 50%.

And, if you view multiple screenshots of these charts taken at different times, they're hard to compare visually, becuase they may end up on different scales.

I think it would be nice if the Normalized CPU % charts were always displayed with a fixed maximum Y axis value of 100%, regardless of the values actually graphed there, and the non-normalized CPU % charts displayed with a fixed maximum Y axis value of ncores*100%. (Where ncores is probably the maximum number of cores that shard had at any time during the displayed time period, so you don't end up with values off the chart in the case where the shard has been downsized and previously had high CPU usage.)