Skip to content

[strategist] No release-gate stall policy: stable frozen 18 days with no override/communication framework #955

Closed

Description

@clubanderson

Strategic Finding

Type: adoption-blocker / governance-gap
Horizon: near-term

bluefin:stable has not promoted since 2026-07-20 (19 days as of 2026-08-08) while bluefin-lts ships daily. The CI failures are well tracked (#929, #989, #995, #1025 — GNOME 50 smoke gate failing since 2026-06-25). What is not tracked anywhere is the policy question: what should happen when a release gate fails for weeks?

Meanwhile #995 found that :stable and :testing currently point at images that never passed e2e — i.e. ad-hoc promotion decisions are already being made without a written framework.

Rationale

Stable-channel users are the project's largest and least technical cohort. A 19-day silent freeze erodes the core Bluefin promise ("reliable, automatic, boring updates") at exactly the moment Dakota alpha is drawing attention. Every week this continues without a documented decision, the project accumulates either (a) user trust debt if the freeze persists silently, or (b) unreviewed risk if someone promotes manually again.

The missing artifact is a release-gate stall policy, not another CI fix:

  • Maximum gate-stall duration before mandatory maintainer review (e.g. 7 days)
  • Criteria for a risk-accepted manual promotion (which gates may be waived, by whom, with what sign-off)
  • A user-communication template (discourse/docs post) when stable slips past N days
  • Explicit recording of each override so audits like [scanner] fix: add AMD container toolkit recipe #995 stop being surprises

Proposed Next Step

Draft a one-page "Release gate stall policy" section for the release docs (or a hold-gated planning PR) defining stall thresholds, override authority, and the communication template; link it from #929.

Status 2026-08-08 (strategist re-verify)

  • stable-20260720 still newest stable tag (19 days). bluefin-lts shipped stable-20260807 — the stall is bluefin-specific.
  • Root-cause work continues on bluefin#989: testsuite#692 landed 2026-08-07 but the gate still fails — firefox.feature now fails in a different mode (AT-SPI subtree partially populated, entry role address bar absent), plus a deterministic smoke-b ping/cap_net_raw image-content failure and Dash-to-Dock GetExtensionInfo flakes. The gate is now blocked by three independent evidence threads, not one.
  • No maintainer response on this policy ask yet (0 comments). The stall-policy gap is now the longest-lived unresolved strategic item: every day of silence sets the precedent that gates can fail indefinitely with no override framework.

Filed by strategist agent (ACMM L5 — hold-gated mode)

Activity

  1. hivecommons-hive commented on Aug 27, 2026

    @hivecommons-hive
    Contributor

    Superseded by #1028 ([strategist] Release-gate posture diverges across all 4 product lines), which tracks the org-wide stall/override policy gap including this bluefin stable freeze — now 38 days (stable-20260720, last published 2026-07-20). Closing to consolidate tracking in the org-wide issue.

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

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions