Changes touching this path

• a contended land gates the merged tree with the harbor let go: a pass whose converge moved the position stops at its gate seam under the new REGATE_DEFERRED with nothing projected, and land_via releases the harbor, runs the merged pass over the tree the converge left in the lane, takes the harbor again and runs the pass again, which projects only when its converge moves nothing and gates again when it moves. every pass a land runs is still whole under the harbor, so the check that landed main did not move is that converge, under the lock that publishes and after the bootstrap catch-up from the checkout and the forge intake, never a tip read like the hoist ADR 0055 records failing. after REGATE_DEFERRALS lost races (2) the next moved converge is gated under the harbor as before, so a stream of lands cannot starve a land. the forge intake is read again each time round: a web ticket write that reached the forge while the harbor was let go is folded, moves the converge and is gated before anything publishes, the #2470 fold check runs on every fold, and a fold made again over an unchanged forge reaches the same merge, since a version id is a hash of its content. the timing: line names harbor-wait each time round and the release as harbor-release, with regate.<row> spans for the gate run outside the harbor. measured on the test fixture with a merged land gate taking 3 s: the harbor was held about 5.35 s before and 2.3 s after over two holds, with land totals of 5.52 and 5.50 s; the real-land number is owed. ADR 0055 is amended, and CONTEXT.md, workflow.md and concurrent.md say so. pinned in 3 new orchestrator tests that watch each run of a declared land gate (the files it saw, and whether a sibling could take the harbor while it ran) as siblings land from the primary. red under named mutations, each restored, over those 3 and the existing regate timing test: the gate always under the harbor (1 passed and 3 failed), the moved check dropped after a deferral (3 and 1), any pass after a deferral published before the defer check (2 and 2), no bound on deferrals (3 and 1), a fold merge whose subject carries the clock (3 and 1). cargo test green, 4826 passed over 144 test result lines with 14 ignored. no deploy owed; lands use it once the primary loot-first is rebuilt; it also carries the comments of review sweep 4 and files its two tickets (mpkswzkl) 7288dd7b · 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.