I love the focus on mobile and smaller format interfaces, quite generally; it's increasingly how I use the projects too!
A) This banner-text-series is clearly impactful, gave me a bit of a jump scare, and got me to read it to find out why. I'm still not sure how I feel about it. ~ Visual effect: Messages that flow smoothly in and out of the reading experience are even nicer. ~ Message: Is there an estimate of the total impact on all readers, as well as total effective fundraising? If there is a very effective compact/delightful banner, and an even more effective large/ambivalent one, is there some internal calculus about the overal impact of running the former for longer vs. the latter for a short period? I'd like to think the best possible messages inspire and delight and draw on positive emotions while raising funds, including for those who don't donate, even if they do not yield the most donations per view.
B) The tracking of whether I've donated, when choosing to show or not show me banners, is definitely lacking. Part of this is that we have taken an overly-paranoid approach to gathering and anonymizing user data. It is entirely possible to cluster users for the purposes of not-continuing-to-show-banners (maintain a dictionary of user-fingerprint-hashes-already-seen, check to see if the current user is in there, don't show banners if they are) without being able to see what pages a given user is viewing.
I wrote more about this here: https://blogs.harvard.edu/sj/2018/07/25/anonymizing-data-on-the-users-of-wik... Please consider doing this; it is really hurting the user-experience of the wiki projects (not only in this instance -- in so many other basic instances of usage stats + testing over time!), for no benefit to anyone.