The free shift roster builder now has a guide, backed by a new bench (eval/callout-churn-bench.mjs), on what happens when someone cannot work after you have posted the week. On its invented cafe, pressing "build" again with the caller off Wednesday changed 4 to 10 other people's shifts, and in three of five cases the caller's shift was still open afterwards. Changing only that shift touched one shift, and on this tight week it found nobody who fit, which a rebuild would have hidden.
Where a city or state has a predictive-scheduling rule, each moved shift may owe a premium. The page prices a 14-day window at one hour per change as an assumption, not as anyone's law. One invented week, so it shows the mechanism, not a benchmark. Everything runs in the browser.