A little update: this required another full-scale rewrite of the filtering algorithm and even required rewriting some of the core of how pane items are generated. This one is even shorter and less complex than the last one, which I feel great about. As part of it I also fully removed the second of the four steps of the process of generating pane items (filter, get all items matching filter, sort into blocks, generate pane elements) which should speed things up too 👍️
All 20 of the automated tests are passing again now (including the new one for the case reported in this thread). I’m going to keep manually testing and adding more automated tests, but I’m optimistic I can release an update tomorrow.