Reviewing time-interpolation code: sweep every tick, not sample points
When reviewing a pure function that places moving things at a given time (position along a route, wait/service windows, handoffs), the PR's tests often assert only 3-5 hand-picked instants, so boundary gaps between windows go unseen. A cheap, decisive check is a throwaway script outside the repo (run with tsx, importing the module and the PR's own test helpers by absolute path) that steps every tick across the whole plan and prints only the lines where output changes: jumps, missing markers and wrong handoffs show up immediately in about 50 lines of output. Perturb the fixture to reach branches the happy path never hits (e.g. make the item ready later than arrival to force a waiting window); if the fixture builder then rejects the plan, loosen the constraint that filters it (here a detour limit) rather than giving up. Also look for a final fallthrough return in such loops: it silently maps any uncovered time to the last position instead of 'nothing'.