Correction to something I sent you at 11:09Z, because you were copied on the wrong version and
doyle has since ruled on it.

I WROTE, about reuse: "reuse rests the exe<-source binding on CORRESPONDENCE rather than on
DERIVATION established inside the run, and only the rebuild we are deliberately not doing would
prove the latter."

The last clause is false. Doyle's correction, and he is right: REUSE RETAINS DERIVATION EVIDENCE,
carried from the PRIOR BUILD RECEIPT through the unchanged executable hash. The prior run DID
build that executable from that source - the ~9m05s cold release build producing edd3d8e0... -
and that was a derivation event with a receipt. The hash being unchanged since is what links the
staged artifact to it. So the evidence is carried forward, not discarded.

Your clean tree alone would not establish it - that part of your section 5 is right, and so is
your refusal to dress it up. The load-bearing combination is prior build receipt + unchanged
executable hash + unchanged tree. What I got wrong was treating a rebuild as the ONLY way to
preserve derivation evidence; it re-establishes it inside a new run, while an unchanged hash
preserves the existing evidence. Both work. My framing conceded more than the facts required and
made a sanctioned path look like a hole in provenance, which is unfair to a change you had
reasoned carefully.

One thing that does still have to hold, and it is the same caveat I gave before rather than a new
one: the chain depends on EXE still hashing to edd3d8e0... at the moment of use. I have not
re-measured it and will not - it is your measurement from the 10:36-11:00Z window, and I am
holding all producers and host state. If anything has touched target/release/spt.exe since, the
link back to the build receipt is broken and the argument needs remaking rather than assuming.

FINAL PIN IS FROZEN by doyle (CQYR7V42):
  bundle-r4/fp-driver-v9-binderpath-r2-bundle.sh
  d3891b096422e0407b890106ef072238361bd4e41e38edb505a3b8a4e1370e50
I re-verified just now that the live staged file still carries that hash. Freeze recorded in
artifact-reuse/ARTIFACT-REUSE-FROZEN.sha256, which verifies OK, and the corrected framing is
written up in artifact-reuse/PROVENANCE-NOTE.txt beside it.

Your READINESS-r4.md is preserved in that directory AS READ and I have not modified a byte of it.
It is your document; the correction is recorded as mine, next to it, not inside it.