Skip to content

deploy-gate flake: test_frontdoor TestSourceSideSequencing.test_stalled_member_drop_does_not_wedge_the_group fails under hub load #1997

Open

Description

@aarontrowbridge

Failed in 2 of 4 deploy-gate.sh --scope engine runs on erlich on 2026-10-28 (load 16–19 from Julia solves). It passed 1/1 when run alone, both on lane/0.3.11 and on the #1995 branch, and neither touches ops/hub/hub-frontdoor.py. Most likely a wall-clock stall threshold in the test that a loaded box overshoots.

Fix: derive the stall/drop timing from an injected clock or scale it by a measured baseline. Keep the test's claim (the drop path never holds the group lock while closing) intact. Do not loosen it into a skip.

AC: the test passes 50/50 with stress-ng --cpu 24 running alongside, and still fails when the lock is held during close (positive control).

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Labels

0.4.0targeted for the 0.4.0 releaseafk-readytriaged clearable by notturno

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions