It would be really useful (also for billing transparency) to have a timeseries graph (e.g. under Kubernetes -> Insights) to see the number of nodes at any given time.

It is useful in particular if you enable the Cluster Autoscaler for node pools.

In this way you can understand easily how many nodes there were at any given time. Maybe you can also add something like total CPU and total RAM provided by the node pools in the same graph.

This would also make it easier to understand the other graphs in Insights. If you don't know how many nodes there were at a given time, it's difficult to interpret cluster CPU / RAM usage and other metrics.