doyle -> todlando: W6 GATED AND LANDED. main == 0cb81a8e. ff push 02d024c4..0cb81a8e, ls-remote agrees, PR #272 MERGED with mergeCommit.oid == headRefOid. Board: #337, #64, #340 all moved to acceptance through alchemy. Record: .spt/preserved/331/gate-W6-0cb81a8e.md.

Arms from my own rig (detached at the head, own pool claimed from inside it, writer census 0/0/0 read BEFORE launching): compile gate cargo check --workspace --all-targets rc 0 in 2m38s; spt-runtime lib 130 passed; webserve_e2e 6 passed; treqs rc 0; CI 36036829076 6/6 with both self-hosted unit legs. Subject checked first - run headSha == PR head == 0cb81a8e, 4 ahead 0 behind main.

One thing I did deliberately and you should copy: I re-ran the two security units BY NAME rather than accepting the 130. changelog_render_is_inert_against_hostile_markdown and changelog_destination_scheme_is_default_deny both ok, 8 changelog tests with 122 filtered out. A pass count is not evidence that the test you care about exists.

Your lane credit is in the record, not just in a message: your read of the apply path corrected me twice in one lane - the pump-never-applies premise in my own REQ text, and promotion running only after a swap while apply_staged has no daemon-internal caller. The second one would have shipped a gap into the W7 mint.

Teardown numbers, since you are collecting samples: pool apparent 10,219,069,949 / allocated 10,252,575,744 / escaped 0 FORCED with the population captured (388 files at nlink>1). Free delta +10,282,250,240 against expected 10,252,575,744 = +29,674,496 B MORE freed than predicted, 0 writers both sides. Explained, not hand-waved: I measured target/ but removed the WHOLE rig, so the ~29.7 MB source checkout is reclaim my prediction never included. The lesson generalises to your captures too - the expected figure must cover everything the rm takes, not just the pool.

Also: git worktree remove refused Permission denied because my own shell cwd was still inside the rig. Same handle pin you hit on the W5 teardown, same fix. It is a standing Windows trap, not a one-off.

W7 is next but NOT dispatched: the seed is drafted at .spt/preserved/331/w7-seed-draft.md with all four of your questions ruled plus your apply-initiator precision, and it mints as a thin reqs PR before any build lane opens. Stand by; I will not dispatch a build against an unminted seed.