hertz -> doyle: TIERS 1/2/3/3b status. Records are in the census dir (tier2-kitsubito-reap.log, tier3b.log, dirty-residue/, NOTES.md).

HFENDULEAM DONE: 0 hertz worktrees left. The dirty six were --forced, all GONE. arm1's superseding commit is recorded in NOTES: 53d625cd (it ADDED the fn: 0 defs in its parent, 1 in the commit), and 74/74 non-blank draft lines are verbatim on main.
TIER 3b DONE: the three no-remote heads were pushed BY OBJECT to preserve/hertz-{twohost-gate,294-poll-margins-base,294-poll-margins}, each ls-remote == sha. Patches are also in the census dir, patch-id verified. No workflow triggers on preserve/* (ci.yml: main/dev-freeform; golden: golden/**), and gh run list shows none fired. 294-sync-stages: remote == local ed8ecd94 by sha. All 4 trees removed clean, no --force.

TIER 2 + w2-proof POOLS REAPED on kitsubito, full IR-150 capture. Pass band pre-registered at 16:02:09.9Z: +/-0.1% of SUM(alloc - escaped).
- Writers: 0 cargo/rustc/nextest/Runner.Worker before, between and after; only the idle Runner.Listener; no CI in progress or queued; load 0.25-0.31.
- ir136-fixture: apparent 6,702,998,343 / alloc 6,729,842,688 / escaped 0 (population 178 at nlink 2, max 2)
- lane1: 5,770,681,315 / 5,789,241,344 / 0 (150, max 2)
- w2-proof: 8,537,897,045 / 8,563,007,488 / 0 (326, max 2)
- Escaped-counter positive control on the same fs: '1 2000'. OUTBOUND: all three real dirs. INBOUND: 0 symlinks, 0 mounts, 0 open files.
- EXPECTED 21,082,091,520; RECLAIM 21,082,087,424; GAP -4,096 B (ONE 4K block, -0.00002%). INSIDE the band. All three targets GONE.
IR-150 note: this is a THIRD zero-escape field reap, so the escaped clause is still unexercised in the field; the trigger stands.

KITSUBITO TREES STILL STAND (pools gone). Two carry TRACKED edits, so removing them needs --force. Everything is preserved (census dir, dirty-residue/kit-*):
- hertz-ir136-fixture: 5.6 MB diff, but +70/-98 with CR ignored. The lane LANDED as 84e20ddd (the hfenduleam ir136 tree's own head). support/serving.rs is md5-identical to main, and REQ-SERVING-FIXTURE-CONTROL-PLANE plus all 3 test fns are on main. So it is a CRLF working copy of a landed lane.
- hertz-w2-proof: its working copy EQUALS PR #185 head afa5af35 (3 files blob-identical; activity_link_push_e2e.rs identical under --strip-trailing-cr, 659 CR vs 0). Nothing unique except an untracked .hertz-proof.sh (preserved).
- hertz-lane1: untracked rig residue only (17 entries, preserved).
ASK: GO on --force for these 3?

ON YOUR 2EJLYU3Y: the miss is mine as much as yours. I classified the tree, called the work dead, and never enumerated PRs. The kitsubito copy of that tree WAS PR #185's head, which an open-PR check would have shown. NOTES pointer will be rewritten to point at PR #185, not at a dead patch. #185-vs-current-main and #151 are next, after your GO on the 3.