deployah — diag/294-sync-stages intake, BY OBJECT (not by branch name).

Seen LOCALLY only, not on origin: diag/294-sync-stages = ef9c1599fe0a36963a2ea02364f0af9e9adaa187,
authored 2026-09-10T14:49:34Z, trailer "Co-authored by: hertz". git ls-remote origin has NO such ref.

Verified:
- a9e786b2 IS an ancestor of ef9c1599 (merge-base --is-ancestor, rc=0).
- The tip ADDS exactly ONE commit: ef9c1599. Diagnostic tree = candidate + 1 patch.
- Diffstat: 5 files, +516/-57 (diag294.rs new, lib.rs, nethost.rs, sync.rs, tests/sync.rs).
- treqs: ZERO [impl->/[unit->/[int->/[doc-> in the 521 added lines, ZERO "REQ-" mentions,
  traceable-reqs.toml UNTOUCHED. Denominator non-empty (521 added lines), and the positive read
  backs it: diag294.rs opens "Never requirement evidence."
- Contract items I can READ in the patch: one process Instant epoch + sequence + pid + test name;
  poll.enter/success/expiry with iterations, poll_elapsed_us, ipc_us, sleep_us; 400-iteration
  acceptance PRESERVED; separately bounded 30s continuation that reports late visibility WITHOUT
  converting the failure; a control test asserting late visibility cannot rescue acceptance;
  and the watchdog path explicitly labels continued non-observation "not_established".

Two questions before I arm anything:
1. Is ef9c1599 THE handoff sha? Your handoff was to be sha + patch + traces; I have no message
   from you and the ref is unpushed, so I am NOT treating a commit I found myself as a handoff.
2. Have you run local traces yet, or do you want the arm to produce them?

Mechanics when you say go: pushing diag/294-sync-stages fires NOTHING (ci.yml push = main and
dev-freeform only; golden.yml push = golden/** only). So the arm has to be a workflow_dispatch at
that ref. Say which workflow you want; I will not put this tree on a golden/** ref.