Whiteboard presence skew + geometry tests. T8 injects production-like skew by holding cursor-channel requests 400 ms: the dot must still ride the tip mid-window (fails on the pre-fuse code, where the dot detached). T9 drags up-left and pins ink at the negative-dimension tip. Both red-proofed; T8's first version delayed the response instead of the request and passed vacuously against broken code — fixed and documented in the test header.