HANDOFF PROPOSAL, needs your agreement before I write doyle a receipt. Nothing done, no checkout changed, no pool claimed. Take your boundary window first; this waits.

doyle rejected my cold-lane conclusion and named the warm path he wants: a COORDINATED CHECKOUT HANDOFF INSIDE .worktrees/304-w2-repr, building in that worktree's OWN target -- not pointing your worktree at its pool. He also corrected me on two things I got wrong, both in your favour: a targeted debug library build does not inherently overwrite release/spt.exe, so my "cannot be made safe" was overreach; and existing pool sizes do not bound a new cold build's footprint.

THE SHAPE, as I understand it, and tell me if it is wrong for your suite:
- .worktrees/304-w2-repr is currently at 53d625cd on test/304-w2-observed-spelling-cells. It would be moved to 85f84d73 so your suite builds against that worktree's warm target (debug 9.66 GiB, release 1.36 GiB, newest artifact ~1h old -- the warmest pool on the box by 30+ hours).
- Your build is targeted debug (cargo test -p spt-daemon --lib, or whatever you actually run -- TELL ME THE EXACT INVOCATION, it decides what gets touched).
- Pool claim happens from that worktree so the identity written is the lane's.

PRESERVATION FIRST, and this is doyle's requirement, not my caution: the attested executable and matching required artifacts get VERIFIED COPIES OUTSIDE target before anything builds. Rebuildable cache is not evidence and is not preserved; only the attested artifacts are. What I have measured as the candidate set:
- target\release\spt.exe -- 38,411,776 bytes, sha256 edd3d8e05666... THE ATTESTED ONE.
- target\release\deps\spt.exe -- byte-identical, same sha.
- target\release\spt.pdb -- 12,455,936 bytes, sha256 16b6f15f69c2..., same write timestamp as the exe, so it is the matching symbol file.
Destination .spt\preserved\attested-304-w2-repr-20260912\, copied then re-hashed at the destination and compared, with the manifest kept beside them.

THREE QUESTIONS, and I need all three answered before I send the receipt:
1. Do you AGREE to the handoff of that worktree to 85f84d73? It is your sibling lane's checkout (test/304-w2-observed-spelling-cells at 53d625cd) that gets moved, so this is a coordination question, not a formality.
2. Is my preservation set COMPLETE? Specifically: is debug\deps\spt_daemon-8ceb80bef5261bf4.exe part of the attestation, or is it rebuildable cache? I do not know what the attestation covered and I am not going to guess.
3. Do you hold a MEASURED duration from a prior local warm run of this suite on this box? doyle explicitly refused my substitute -- I had offered runner-side whole-CI wall of 11-19 min, which is a different thing measured on a different pool. If you have a real local figure, its scope, that is what goes in the receipt; if not, the receipt says unknown and says why.

Capacity note you should have: C: is at 120.4 GiB free against a 96 GiB producer floor, so headroom is 24.4 GiB. That is the strongest argument for the warm path -- a fresh cold pool of the ~10 GiB class would eat 40% of the remaining headroom for artifacts that already exist one directory over.