Arm 2 (bare HEAD a9e786b2, twohost=false, run 34474627303) status at 12:38Z.

JOBS. traceability success 12:04:15. changes success 12:04:24. n1-gate(Linux,kitsubito) success
12:05:58. test(Linux,kitsubito) COMPLETED SUCCESS 12:06:00 -> 12:35:06. test(Windows,hfenduleam)
STILL in_progress since 12:04:26. n1-gate(Windows,hfenduleam) still queued (its runner is under the
Windows test leg). Run-level conclusion still empty; I score from jobs, not the run status field --
that field read "queued" at 12:11 while two jobs were in_progress and three were terminal.

LINUX LOG PULLED AND PRESERVED (682,227 bytes) at
.spt/preserved/arm2-bare-head-34474627303-20260910/test-linux-102862459272.log

MEASUREMENTS, Linux leg only.
Control first: grep -c on the Summary line = 2, and the two carry DIFFERENT populations (3360 and
223), i.e. the two phases, not the repeated-population VOID shape.
  Phase A Summary [  48.570s] 3360 tests run: 3360 passed (1 leaky), 1 skipped
  Phase B Summary [1172.538s]  223 tests run:  223 passed (1 slow), 0 skipped
Failure-class rows with a suffix allowed (FAIL/TIMEOUT/SIGSEGV/ABORT): ZERO. Occurrences of the
panic string: ZERO. That reconciles with both Summary counts.
Both phases ran with --no-fail-fast (read from the two step command lines), so neither Summary is a
prefix of an aborted population.

TARGET CELLS, both executed on Linux:
  PASS [ 1.408s] (  27/3360) spt::contract_e2e live_agent_lifecycle_e2e
  PASS [ 1.468s] (1745/3360) spt-daemon::sync two_tier_sync_lands_and_gate_refuses_server_side
Neither carries a SLOW row here. I am NOT comparing 1.468s against arm 1's 91.375s: that one is
Windows, this is Linux, different box and different tree.

ALSO: arm 1's Linux red cell executed and passed here --
  PASS [ 3.629s] (  49/3360) spt::io_events_undriven_kinds_e2e
    relay_backlog_and_live_deliveries_stamp_once_even_after_hook_poll
Scope limit I have to state with it: that assert compares (kind, author) and BOTH expected rows are
("MSG_IN", Some("alice")), so two BACKLOG rows satisfy it. This green does not establish that both
relay arms published. n=1, different tree from the arm that reddened it.

Only other non-PASS rows in the leg: one LEAK [0.208s] on
spt-daemon livehost::tests::legacy_psyche_sweep_guard_is_id_specific_and_fail_safe (a leak row is
not a failure) and one SLOW [> 60.000s] on
spt-daemon::transport_death_eof transport_death_without_a_fin_surfaces_eof_to_the_serving_worker.

WHAT THIS DOES NOT DO: it classifies nothing. The verdict of record stays golden 34445961595
attempt 1 @ a9e786b2 WITH twohost, conclusion failure, carried by test(Windows). twohost=false arms
are not release-gating. Classification is doyle's.

Windows leg pending; I will pull and report its log the same way when it goes terminal.