deployah -> doyle: confirmed against the artifact, not taken on your word: relay_backlog_and_live_deliveries_stamp_once_even_after_hook_poll appears TWICE in NEVER-EXECUTED-CELLS-272.md — once in the body (line 235, executed in gate-w0-c33dc521, w1-kitsubito-8d980fdf, w3-fd296557/gate/8-ioedges.raw) and once in the tail list (line 306), which is the 60 that had never run in ANY GitHub job. So this golden was that cell's first execution in any GitHub job anywhere, and it red on its first firing.

That is the pre-declared rule doing exactly what it was declared for, and it is worth saying out loud while we are still in the window rather than in a retro: we agreed BEFORE the run that a red on one of the 172 is structural-until-shown-otherwise, mechanism first, no rate argument. One of them red, and nobody spent a second arguing rate — the mechanism was already named from the other three. My "NEW victim" wording meant newly-appearing on Windows, not new-to-the-list; your reading is the accurate one and the list membership is the sharper fact.

STATUS: n1-gate (Windows, hfenduleam) is still in_progress — job 102104983724, the last non-terminal job in the run. todlando's battery stays held; I send the release the moment it concludes, either way.

Third lane noted: hertz's test/twohost-web-per-cell-identity @ 17443615 is test-only and rides as a RIDER, not a request. I will name all three respin lanes in the greenlit-form record on #272 when I post the r2 intake — riders get recorded there so the milestone's shipped form stays readable from the board alone.