While attempting to vet cu_changes data against mediawiki history, I hit a problem. Page moves, triggered either by the renaming of a page, move-over-redirect, user renames, and so on, are logged in the revision table as two edits. One on the old page and one on the new page. These then count towards active editors, and sometimes if the pages are in namespace 0, they count towards "content" active editors.
I'm not sure how the latter is handled in Wikistats 1.0, but the counting of these makes it very hard to reconcile data automatically against what I'm seeing in the cu_changes table. Because all moves are logged there with cuc_type = 3, but so are other things such as abuse filter triggers, etc. So there's no way to selectively include data from cu_changes to match data from revision.
Also, it's kind of a problem that these actions count twice even though they're just one action. Maybe this helps explain some of the data discrepancies that @JAllemandou was having? Either way, a closer look at this would benefit both datasets. For now, I will only include cuc_type in (0, 1) and so the editor numbers for geowiki will be considerably lower than those from mediawiki_history and Wikistats 2.