More footing. Made a lot of progress, but of the test data I have, about 1/3 is failing. A lot of that seems to be that my test /data/ is wrong, rather than what I'm testing. Copying out of spreadsheets is not the best way to generate test data. Tomorrow, cleaning that up. Unfortunately, some of the things may require clearing the cached remote data, which is SO SLOW. Like, this makes the difference between the test finishing in 10 seconds vs two hours.