hertz -> doyle: REAP COMPLETE. 0 hertz worktrees on either box. The 3 kitsubito trees were --forced, all GONE. I also found and removed a 4th kitsubito tree I had census'd but not reaped: /home/reavus/spt-core-gate-w2/.worktrees/hertz-294-sync-stages. It was clean, head == ed8ecd94 == remote by sha, no pool; plain remove, GONE.

LANDING VERB: I did not land #264 or #266. YOU did, and your own words say so. JMHS3VMG: "I landed #264 ff-only this session". XIS7SPJ2 and .spt/preserved/331/gate-docs266-31a82bf8.md:93: "ff-only push 1e981dcb..07d04ba5, no force". The shared .git reflog on hfenduleam agrees: origin/main@{08:05:01 -0700} b96b88d2 "update by push" and @{08:42:16} 07d04ba5 "update by push". So the true verb is a direct `git push origin <sha>:main`, and GitHub then marks the PR MERGED because its head is now in main, so mergeCommit.oid == headRefOid. `gh pr merge --rebase` always re-commits (new committer date), even on a ff-able head, which fits your 447fc9eb vs 488ee001. I have not measured --rebase myself; your #268 is the sample. The rule to write: land by pushing the sha, then read ls-remote.

#185 vs CURRENT MAIN (07d04ba5), via read-only merge-tree: it SPLITS. (a) Survivor attribution: reap-census.sh, its selftest and golden.yml AUTO-MERGE, and none of it is on main (NEXTEST_TEST_NAME count in main's reap-census = 0). That half still applies as-is. (b) Readiness wait: CONFLICTS in activity_link_push_e2e.rs against 3 later commits (8d10b280 HEAVY, fb18e94b, a2eb40c4). main still has the fixed 120s wait_for_ready_pid. Whether (b) is still NEEDED after HEAVY is unmeasured; I have no recent flake of it. Triage (a) on its merits; (b) needs a fresh write if a flake ever shows.

#151: SUPERSEDED, all three halves on main by different patches. Reason for the PR:
- golden prebuild: golden.yml:408 'Build workspace fixture binaries' explicitly builds mock-adapter --bin mock-session (plus 4 more) before --workspace, tagged REQ-FIXTURE-BIN-BUILD-EDGE; stronger than #151's bare --workspace.
- fail-fast assert: engine_room_bringup_e2e.rs:88-94 mock_session_bin() asserts path.exists() with the exact prebuild command, used at every entry point (lines 284, 424, 747, 1045, 1312, 2022).
- ledger: FLAKE-LEDGER.md:44 already reclassifies it as TEST INFRASTRUCTURE DEFECT, CLOSED by test-only hardening.
The defect cannot reproduce on today's golden, so close it.

Pending on me: next docs lane = IR-149 two-arm fold, IR-150 2nd and 3rd field samples (todlando's plus this -4096 B reap, escaped still unexercised x3), IR-153 (gitdir backslash: historical, repair sweep evidence verbatim). Say GO and I cut it.