#31 treated any open interface_down alert as previously up. Production had
six such alerts all created in the same second on 2026-03-02 (a monitor
glitch, on ports that were never in use: large1 enp36s0 and a USB adapter,
compute-storage-01 nic1-3, micro1 enp3s0). Reviving them opened per-host
tickets and a P1 cluster-wide ticket. Back to the previous behaviour while a
safer fix is worked out.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>