todlando -- FOLD-1 EXECUTION, START 2026-09-11T12:22Z. SUBJECT 61d3459fd6db5be0e66951b9f6d2a05822865518, lane fold/304-w2-admission, worktree .worktrees/304-fold-admission, pool .worktrees/304-fold-admission/target (label todlando-304-fold), base 697eb398. Tree CLEAN. First act of the window.

ADMISSION: doyle's release at 12:16Z plus my own reader, positive control TRUE (a live cargo.exe seen by the same name filter that reports the zero), ZERO cargo/rustc/cargo-nextest/link/Runner.Worker. Free 142.4 GiB, floor 96 PASS. My own two builds are the only things that have touched this pool.

COMPILE-GATE AT THIS SHA: `cargo test -p spt-daemon --lib --no-run` EXIT 0, ZERO warnings, direct capture.
  The first attempt exited 0 WITH ONE WARNING -- `spec_satisfied_by` was pub(super) while `Rule` is private, a private-interface leak. I narrowed the seam to module-private rather than widening `Rule`: its only callers are `decide`, `reconcile` and the cells, all in this module. Zero-warning discipline holds and the seam did not grow a public surface to satisfy a lint.

TRACEABILITY, and the instrument MATCHES here for the first time in this milestone: this tree pins WANT=0.4.1 in ci.yml and my CLI is 0.4.1. No IR-104 mismatch to declare on this lane.
  REQ-WEB-LAN-BOOTSTRAP-FIREWALL: required [doc, impl, unit], stages +doc +impl +unit -int. The int clause is folded in and stays OFF.
  check exits 1 with NINE [must] misplaced_tag. All nine are code-stage tags sitting in docs/ PROSE (F-035-CONFLICT, PUMP-W2-DESIGN, INFRA-REGISTER, DEBUG-CONVERGE-PLAN) -- files I have not touched, zero naming bootstrap_firewall. Inherited, not mine.
  (Note the population difference from my W-1 lane's 312: this tree carries [placement] module_banner = "accept" from 33936943, the unratified loosening I flagged. Same checker, different policy, different count -- which is exactly why I am naming the tree the number came from.)

CAP 600 s. Arms, mutations first, each reverted immediately and the revert proven by oid AND byte shape (the file is CRLF here; anchors are rebuilt from the file in hand and translated before counting).

  F1 u1-R   mutation: desired_spec ignores its bound_port argument and uses a fixed 5470.
            PREDICT RED in the_spec_is_built_for_the_bound_port_and_a_stale_port_rule_does_not_satisfy_it,
            at the `want.port == BOUND` assertion, left 5470 right 56025.
  F2 u2(ii)-R mutation: DESIRED_PROGRAM flipped to true.
            PREDICT RED in a_program_bearing_rule_does_not_satisfy_a_spec_that_wants_no_program_filter,
            at `want.program.is_none()` -- the FIRST assertion in that cell, so the failure names the
            policy constant rather than the predicate. If it instead fails later at the
            spec_satisfied_by line, my reading of the cell's order is wrong and I will say so.
  F3 hertz's SIX cells by exact name, unchanged by this commit. PREDICT 6 selected, 6 passed.
            ZERO-MATCH GATE: an arm not reading exactly 6 tests run is a NON-MEASUREMENT, not a pass.
  F4 my FOUR new cells by exact name. PREDICT 4 selected, 4 passed. Same zero-match gate at 4.
  F5 unfiltered -p spt-daemon --lib. PREDICT run total == `cargo nextest list -p spt-daemon --lib` total
            at THIS sha, 0 failed. Both numbers reported side by side, both carrying 61d3459f.
            I file NO absolute number in advance: the W-1 lane's 981 was measured on a different tree and
            has no standing here.
  F6 no SPT_POOL_FOREIGN, no lane-identity refusal.

WHAT WOULD MAKE ME WRONG IN A WAY THAT MATTERS: if F1 or F2 comes back GREEN, the cell is vacuous the way u2's extract_rule_programs assertion was, and the seam would be decoration rather than evidence. That is the whole point of running mutations first.

END carries: every number with its sha, both mutation arms quoted at their assertion, the two filtered counts against their gates, list total vs run total side by side, the leaky set, each prediction against its result, the revert oids and byte shapes, and the post-census.