Changes touching this path

• the land test gate names the units still running once a minute while they run, where a unit that hung was not named until it ended: Spawner::capture_each and spawn::pool take an interval and a still callback that hears the running indices on the calling thread, the one that prints each unit block, on a deadline that arriving results do not put off, and the gate prints >>> still running after Ns: with each unit named as its Running heading names it, every STILL_RUNNING_EVERY, 60 s. the pool still owns its receiving end inside the scope, so a panic on the calling thread stops the workers taking units and waits for the ones they took; an interrupt was checked to end the children with the gate, a break to its process group ending all three probe children, while a kill leaves them running, and the pool doc says both. the test gate docs say largest where they said costliest or heaviest, CONTEXT.md gives the doctest retry its condition that no binary failed, TEST_JOBS_CAP says it is a measured cap and not derived from the machine, and DOCTEST_WEIGHT carries a figure of its own, 3.6 and 4.9 s alone. the grant cache docs say a move that stops at a write that did not land is filed and started again by the next unlock, and that addressCount can count an object twice for the session, at the count, at the session bar that reads it and in CONTEXT.md. red under named mutations, each restored: still never called (1 failed and 13 passed), the deadline put off by each result (1 and 13), finished units named (1 and 13), a refused send ignored (1 and 13), the receiving end held outside the scope as the first draft had it (1 and 13), and the line naming the step rather than the unit (1 and 0). cargo test green, 4900 passed over 144 test result lines with 15 ignored, and the site gate green at 998 passed (qnzmnrnz) 65ceec99 · dbf3dbe6…

Renames are not followed. loot's tree maps a path to an address, so a rename is a delete and an add. This list is the history of the name, not of the bytes.