ci(1979): arm the stale sweep — .github joins the wave, and the flip lands last

[LukasWodka]

The last PR of the backend#1979 wave, and the last exit-criterion work for backend#1680.

Two things, in one PR because they cannot be separated

1. This repo gets its caller. It was deliberately left out of #288 (Bugbot's
call): exempt plus a caller on the audited branch is itself a finding
(caller-drift.py:2260), so shipping the caller alongside the reusable would have
gone green on a repo the audit had been told to skip.

2. All 17 migration entries flip exempt → required, and the
stale_backlog_migration_in_flight anchor is deleted rather than left behind.
Its own text asked for that:

if this anchor is still here after that, it has become the same finding as
wip_limit_check_has_no_callers — a reusable that shipped and was never wired up,
with a written excuse. Treat it as a defect then, not a justification.

What is deliberately NOT flipped

release-train and rfcs keep their exemptions, because those rest on reasons that
are still true and are not the migration:

repo	why it stays exempt
release-train	enrolling the repo that drives promotions into an unattended closing sweep is a decision someone should take deliberately. Still UNDECIDED, and the note says so rather than inventing a basis.
rfcs	it holds decision documents, not a backlog. An RFC that has sat untouched is not stale — it is a decision nobody has taken, which is what the board card is for (.github#138).

Ordering — the reason this is a separate PR

This merges LAST. Flipping to required before a repo's caller lands would
redden the org audit for every repo still waiting: the drift window made to look
permanent, which is exactly how a tier teaches people to ignore it.

step	state
1. .github promotes develop → staging → main, carrying #288's reusable	pending — main still has the old 1174-byte copy, develop is 4 ahead
2. the 16 caller PRs merge	open as drafts, one per repo
3. this PR	draft until 1 and 2 are done

Verification

• 17 entries required, 2 deliberately exempt — asserted by parsing the
  inventory and printing every repo's resolved value, not by counting diff lines
• stale_backlog_migration_in_flight: 0 occurrences remaining
• make check green — 41/41 house-rules cases
• caller-drift-selftest.py — 197 pass, 0 fail
• actionlint clean on the new caller

🤖 Generated with Claude Code

---

Note

Medium Risk
This is the org contract the caller-drift audit enforces. Flipping many repos to required will redden the fleet audit if any caller is missing or not yet on the audited branch.

Overview
Closes the backend#1979 migration in repo-inventory.yml by flipping stale-backlog.yml from exempt to required on the repos that were waiting on caller PRs (including .github), and deletes the stale_backlog_migration_in_flight shared reason so it cannot linger as a written excuse.

release-train and rfcs stay exempt on their existing, still-true reasons (undecided enrollment of the promotion driver; RFCs are decisions, not a backlog).

^{Reviewed by Cursor Bugbot for commit a8c5174. Bugbot is set up for automated code reviews on this repo. Configure here.}

[LukasWodka]

bugbot run

[LukasWodka]

Split: .github's own caller moved to #307, because bundling it here could never go green

All 16 other callers are merged, and this PR's flip cleared 16 of the 17 findings. The seventeenth could not clear, and re-running would not have helped:

1 drift finding(s):
- .github: MISSING required caller for stale-backlog.yml on develop.

Fresh run 32595845377 at 20:09, after the wave landed.

The mechanism is a deadlock, not a flake. The audit reads caller presence from a repo's audited branch via the API — --source-dir covers only copies and reusables — so for .github itself the caller does not exist on develop until it merges. With the entry flipped to required and the caller only in this PR's diff, the audit must report it missing. And gate is a required context on develop. So the PR that adds the caller cannot merge until the caller is added.

The caller now lands in #307, which touches no contract file and therefore does not trigger conformance-gate at all (GUARDED lists repo-inventory.yml, caller-drift.py, its selftest, caller-drift.yml, conformance-gate.yml — none of them in #307).

On the deliberate bundling this reverses

The caller's header said it shipped with the flip "deliberately", because exempt plus a caller on the audited branch is itself a finding (caller-drift.py:2260).

That finding is real and will appear on .github between #307 merging and this PR merging. But it is a transient window closed by the very next merge — the same window the whole fleet just went through, bounded by the stale_backlog_migration_in_flight anchor — whereas the bundling it was avoiding produces a PR that can never be merged at all. I rewrote that header in #307 rather than leave a rationale describing a plan no longer being followed.

Splitting also restores this file's own governing rule, stated in repo-inventory.yml's header and followed for all 16 other repos: land the caller first, flip the entry after. This repo was the one exception, and the exception is what jammed.

Order

1. ci(1979): land .github's own stale-sweep caller, ahead of the flip #307 — .github's caller onto develop
2. this PR — the 17-entry flip + anchor deletion; its audit should then be green, every one of the 17 callers existing
3. fix(2242): say why rfcs is exempt, accurately, and stop one anchor describing two repos #306 unblocks as collateral, and backend#1979 closes

Also merged develop in (branch was 9 behind), which fixed a genuine quality / pipefail early-close failure — that check was judging pre-#2264 shell.

[LukasWodka]

bugbot run

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit a8c5174. Configure here.}

[cursor]

✅ Bugbot reviewed your changes and found no new issues!

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit a8c5174. Configure here.}

[saadqbal]

Not approving, because gate and audit are both red and a failing required check is a hard no — but the red here is diagnosed, not mysterious, and it isn't this PR's fault.

The audit run reports one finding:

- .github: MISSING required caller for stale-backlog.yml on develop.

and gate then refuses on caller-drift.yml concluded 'failure' ... A contract change must not merge on a failed or skipped audit. That's the deadlock #307 was split out to break: the audit reads caller presence from the audited branch, so .github's own caller does not exist on develop until #307 merges, and a PR that bundles the exempt → required flip with the caller can therefore never go green.

So the sequence is #307 first, then re-run these two checks here. I've approved #307 on the strength of the same log. Nothing to change in this diff on that account — when it goes green I'll review the flip itself on its merits.

Flagging one thing so it isn't a surprise: between #307 merging and this one merging, .github's audit is expected to be red for the opposite reason — exempt in the inventory plus a caller present on the audited branch is itself a finding (caller-drift.py). That window is real and closes on this merge, which is the trade #307's comment argues for. Worth not reading that red as a regression when it appears.

[saqlainsyed007]

Verified against the tree at head (a8c5174), not the description:

• stale_backlog_migration_in_flight anchor: 0 occurrences remaining — deleted as its own text asked, not left behind as a written excuse.
• 17 entries flipped to stale-backlog.yml: required, 2 kept exempt — and the two are exactly the ones the description names: *stale_backlog_exemption_needs_redeciding (release-train, enrolment of the promotion driver still undecided) and *stale_backlog_not_a_backlog (rfcs, decisions not a backlog). Neither rests on the migration, so neither should flip. Correct.
• Decisive check: audit and gate are green at this head. caller-drift reddens the moment a repo flips to required without its caller present, so the green audit is the derived proof that every flipped entry — .github included — has a real caller on the audited branch. This is exactly the outage the staged ordering exists to prevent, and it isn't tripped.

Bugbot pass, 3 threads all resolved, mergeable. BLOCKED is the ordering/branch-protection gate (this lands last), not a conflict — that's the train's concern, not a code issue. LGTM.

[LukasWodka]

/fr-pass