I spaced and realized I had already posted about this (thanks so much, brain) to which you had already given some ideas. The original post actually has a more correct power graph - there are two large >1100W sustained peaks, a ~30m period of ~150w, then a final peak of around >1100W for about 15 minutes. As soon as that final peak drops off to below 50w, that’s the period I want to catch.
I’m not sure the logic you suggested here would hold true in this case, but it may require some experimentation. Thanks for the new tool to try.