Thanks for those ideas. As far as I know, the cumulative totals don't get reset unless the router is actually rebooted, so missing data shouldn't be too much of a problem so long as the data can be read close to the turn of the month. Individual days may be wrong, but the monthly totals will be fairly accurate.
I was planning on collecting and saving the data with every sample, and analysing the figures as part of the normal post-sampling processes. Where there are discontinuities in the timeline I could, as you suggest, display the daily data in a different colour.