Every issue closed on this repository before it moved to Backlog.md on 2026-08-14. Six issues,
all COMPLETED, #6 through #46.
They were not imported as Done tasks, deliberately. Backlog IDs follow creation order, so
importing them would create a second ID space over the same history — a GHC-000N that can never be
made to match the #NN already cited in commit messages, in the other repos’ issues, and in the
Refs/Closes footers of published history. The #NN space stays the only one. This index keeps
the history readable from the checkout alone, and costs one file.
Where the bodies are
The issues were deleted from GitHub after this archive was pushed, so gh issue view <N> no longer
works. The full bodies, acceptance criteria and completion comments live in the repo:
jq -r '.[] | select(.number == 32) | .title, .body, (.comments[].body)' archive/github-issues-2026-08-14.jsonarchive/README.md documents what was captured, how comment completeness was verified against the
REST API, and the identifier sweep. Nothing in the dump required redaction; the README carries the
evidence rather than asserting it.
The index
| # | Title | Closed | Resulting SHA |
|---|---|---|---|
| 6 | Roll out OpenSSF Scorecard across the public fleet (reusable workflow + badges) | 2026-07-03 | 8537db9 (workflow), released as 8718898 v1.4.0 |
| 13 | Enable Renovate platform automerge on the hub repo (match fleet) | 2026-07-03 | 4369add — plus API-only changes with no git trace, see below |
| 14 | Renovate action-update merges don’t trigger a release-please release | 2026-07-03 | de5c667 |
| 15 | release-please: un-hide chore so already-merged action bumps land in the release PR | 2026-07-03 | d365eee |
| 32 | container-publish: an all-numeric short SHA makes the main chart version invalid semver | 2026-07-24 | 3b135db |
| 46 | fix(ci): verify actionlint downloads across shared workflows | 2026-08-12 | 328bc72 |
The three that still constrain present-day work
Most of the above is finished history. Three carry decisions that a future change can silently undo, and the reasoning is in the Wave operating model doc rather than repeated here.
#13 is only half in git. allow_auto_merge, delete_branch_on_merge and the main branch
ruleset (requiring the ci-success check, ruleset id 18475192) were applied through the API and
exist only in GitHub’s state. Reading the diff will not tell you they are there, and reading the diff
is how a future session concludes they are not.
#14 and #15 together are why chore is releasing here. chore was un-hidden in
release-please-config.json so already-merged action bumps could be swept into the open release PR,
and this repo’s renovate.json overrides the fleet default to commit dependency updates as
build(deps). The consequence — routine chores now cut patch releases on this hub — is intended,
because consumers pin the reusables to release SHAs. Re-hiding chore to “tidy the changelog” would
reintroduce the exact bug #14 describes.
#32 and #46 are the two live proofs that a hub defect is a fleet defect. Both were one-repo
fixes with a dozen-plus consumers, both reaching those consumers only as each bumps its SHA pin.
Not in this index
Pull requests, which gh issue list excludes and which the git history already records. And #3,
Renovate’s Dependency Dashboard — still open, deliberately kept, and recreated on every Renovate run
regardless.