origin/main fails its own gate ratchet at its own tip: no session can reach a green preflight

Row: ENG-PREFLIGHT-COMPILES

origin/main fails its own gate ratchet at its own tip. This is not a branch-local
problem and it is reproducible from a pristine export:

T=$(mktemp -d); git archive origin/main | tar -x -C "$T"
cd "$T" && python3 scripts/check-gate-commands.py --check; echo "rc=$?"

ERROR: these baseline rows left the gated population: ENG-PREFLIGHT-COMPILES.
If that is a legitimate record edit, re-pin RUNNABLE_BASELINE in the SAME change,
naming each row and the reason.
rc=1

scripts/agent-preflight.sh runs exactly this invocation (agent-preflight.sh:371,
the claim-view|check-gate-commands arm passes --check), so every session that
runs preflight on any branch inherits this red and cannot reach a green preflight,
whatever its own change does. tests/scripts/test_check_gate_commands.py fails for
the same underlying reason, because its ratchet test reads the live tree:

FAIL: test_adding_a_row_that_is_not_runnable_also_breaks_the_pin
AssertionError: Items in the first set but not the second: 'ENG-PREFLIGHT-COMPILES'

ENG-PREFLIGHT-COMPILES is present in RUNNABLE_BASELINE
(scripts/check-gate-commands.py:466-473, added 2026-08-31 with a "GROWTH, and
earned" note), so the pin was written; what the checker reports is that the row no
longer RESOLVES into the gated population it was pinned against. One of the two
moved after the other was written.

Not fixed here on purpose. I found this while landing VT-CPU-ELEM-SURVEY
(#2416), whose branch reproduces it only because it merged origin/main. Re-pinning
a baseline I do not own, to make my own preflight green, is precisely the "never
weaken a checker to make a transition pass" failure the same output warns about, and
the checker's own message asks for the re-pin "in the SAME change" — which is the
change that moved the row, not mine. The row that owns the pin owns the repair, and
it should say which side moved rather than re-pinning to whatever the tree now says.

Found while working #2416; that branch records the same finding in its merge commit
message rather than acting on it.

Run the pristine-export command from the issue, then read scripts/agent-preflight.sh:371, scripts/check-gate-commands.py:466-473, and tests/scripts/test_check_gate_commands.py. Trace why ENG-PREFLIGHT-COMPILES no longer resolves into the gated population and identify the change that moved it; done means the checker and ratchet test pass without weakening the gate, with the owning baseline change documenting the reason.