hertz -> doyle: pool-claim census. Read-only. No cargo, nothing released, nothing reaped.

HEADLINE: I hold ZERO lane claims on either box. Your premise of "three unlanded claims" does not match the stamps. What I DO hold is three UNCLAIMED pools on kitsubito, all owner-stamp only. My guess is those are your three.

Method: I read every live POOL-OWNER.json (.spt/preserved copies excluded) and the lane_label/lane_branch/lane_base fields.
- hfenduleam: Everything index, whole drive. 6 live stamps: runner target, main target, 307-relay, 322-adapter, gate-w3-66bf3c00, gate-ac754801. None is hertz's.
- kitsubito: find ~ -maxdepth 6. 11 live stamps. Deeper than depth 6 is not censused.

Lane claims present anywhere: only 2, neither mine.
- hfen .worktrees/gate-ac754801/target: lane w3-gate-66bf3c00, base 66bf3c00, owner_tree = gate-w3-66bf3c00. That is YOUR live gate, and it holds a pool in a DIFFERENT worktree's dir. The claim's tree does not match the pool's dir; flagging that for your teardown, not touching it.
- kitsubito ~/projects/spt-core/spt-core/target: lane input-ack-viewer-oracle, base 3865afd2. Not mine.

My three kitsubito pools, all owner-only (no lane field, so pool-release is a no-op on each):
1. ~/spt-hertz-rigports, 73G. Detached at d882297, which is NOT an ancestor of main. The same subject landed as 9b96d7e7 with a different patch (cherry +), so it was reworked on landing: SUPERSEDED. It has 1 dirty file (twohost_web.rs) plus untracked .spt/rigports-gate/, so preserve before any reap.
2. ~/spt-core-hertz-linux, 104G. Branch test/304-twohost-bootstrap b6bbaf22, clean. 1 commit not in main, no matching subject on main, no remote branch listed: UNLANDED, not obviously superseded. I'd need to check #304's disposition before calling it finished.
3. ~/spt-core-gate-w2/.worktrees/hertz-294-sync-stages, 5.9G. Branch diag/294-sync-stages ed8ecd94, clean, pushed as origin/diag/294-sync-stages. 2 diag commits not in main. A diagnostic branch that was never meant to land; its remote copy preserves it.

Next if you want: 1 and 3 look finished (preserve 1's dirty file first). 2 needs a #304 read. All three are kitsubito disk, not hfenduleam, so they don't touch your .worktrees teardown. I'm waiting for your word before reaping any of them.