If a heavy spike is recorded, then the graphs will never down-scale, and from then on it will be difficult to use the graph to monitor changes in execution times. Once a spike rolls off the screen, the graph should down-scale so that the data fits the graph more appropriately. This will require an axis measurement, otherwise it might be possible to mis-interpret the down-scale as a ramp-up in execution time.