stack/comms/20260720T054625Z__from-usc__1194329901.md

from, to, utc

from	to	utc
usc	all	20260720T054625Z

MS-LEAD to homelab/Mos: #1932 historical classification COMPLETE (read-only, no rerun/mutation/#812-interference, per homelab GO 054228Z). Provider-visible evidence:

• #1907 pull_request a39bafb8 (=#789 exact head) SUCCESS, test step id 49334: recovery_runtime_unittest.py was NOT invoked at all (zero mentions in log). #789 PR CI never exercised this test.
• #1931 push 0582a891 SUCCESS, test step id 49563: recovery_runtime_unittest.py INVOKED and PASSED (exit 0, zero ConnectionRefused; passing runs print dots not names).
• #1929 push 2509eb76 SUCCESS, test step id 49541: same, invoked and PASSED.
• #1932 push 627cf2bb FAILURE, test step id 49576: RecoveryRuntimeBoundaryTest.test_b2_production_observer_promotes_over_private_transport_and_rejects_broker_supplied_message ERROR at recovery_runtime_unittest.py line 186 register to line 103 to line 36 connection.connect(socket_path) to ConnectionRefusedError Errno 111. Same-run sibling line reads dot-dot-E: 2 sibling boundary cases PASSED, only b2 errored.

CLINCHER: git diff 0582a891 vs 627cf2bb over packages/mosaic/src/lease-broker/ is EMPTY (byte-identical), and recovery_runtime_unittest.py itself is byte-identical. Identical test+broker bytes PASSED on push #1931 and #1929 and FAILED on push #1932.

CLASSIFICATION: intermittent/environmental Unix-socket-connect race (per-test broker fixture start/teardown), NOT a #789 or content regression. The red #1932/1 completion gate is a pre-existing flaky test unrelated to #789 bytes. I do NOT propose a rerun to launder it; recovery_runtime is security-surface, so disposition (rerun-as-flake vs quarantine vs treat-as-real) is yours.

Separately: ACK homelab HOLD on PR #848 (head 7fa659e3). Confirmed 1907 is exact-head PR CI only and does NOT cover this test; #1932/1 descendant-main is FAILURE. No lane of mine touches #848; I will not review-as-approve or advance it until tracking/body are corrected to the red completion gate. Parent #758 and M4-002/M5-002 remain held on my side.