Currently, sparklines are sorted alphabetically. However, sparklines are often used to investigate regressions in summary timeseries or top-level metrics. That task would be easier if sparklines that were numerically more similar to the main chart (around the selected revision) were on top.