The Reclaimed Bytes graph . . . has issues. There are times where lots of reclaim events occur in short periods, resulting in columns of squares on the graph, but we lose the ability to tell the graphs apart. See: https://imgur.com/a/nRNnFNX
That graph is done using an exponentially decayed weighted moving average. (Big fancy words, but it's perhaps the most simple moving average to implement... y(x+1) = y(x)*.9 + measurement*.1 -- except the requirements here would make it a bit more complicated (log and exponent to handle variable time intervals). In the case of the graph I linked, it's recording single events. It only ever sees single documents to process, but it may see documents with 0ms gaps or 10 hour gaps between them. Rather than a row of dots at zero or 1 document, the weighted average makes the workload much much more clear, which is something I think might benefit the reclaimed bytes graph, and I'm wondering if such a graph may indicate one of the activity spikes on May 7th is noticeably different from the others.
Then again, too many different ways to show the same data may confuse users . . .