[Ticket]: Spike — the GitLab multi-forge epic cites ADR-0077/0078, which do not exist; decide: resurrect the ADR or drop the epic

### Type

Spike

### Summary

The GitLab multi-forge epic **#588** and its tickets **#589**, **#590**, **#591** (plus the now-closed `ai-helm-values`#300) all cite **ADR-0077** and **ADR-0078** as their source of truth. Neither number means what those tickets say on `main`:

| Cited as | What the ticket claims it is | What it actually is on `main` |
|---|---|---|
| ADR-0077 | "adopt IRS+SPI for the OpenCode CLI" (#591) / "local-dev is already forge-neutral" (#588) | `0077-my-usage-dashboard.md` **and** `0077-phoenix-style-chat-dashboards.md` — two dashboard ADRs (see #1073) |
| ADR-0078 | "GitLab CI via forge OIDC" / "multi-forge repo binding, GitLab OIDC" (#589, #590, #591) | `0078-per-user-span-attribution-for-chat-content.md` — already **Superseded by ADR-0079** |

**There is no GitLab or multi-forge ADR anywhere in `docs/adr/`.** A case-insensitive grep for "gitlab" across all 136 ADRs returns only two passing mentions — ADR-0007 ("a GitLab CI component is a possible follow-up") and ADR-0009 ("GitLab support deferred") — neither of which is a decision record for this epic.

#588's own AI Usage Declaration says the assistant "drafted ADR-0078 + this decomposition". That draft appears never to have merged, and the number was subsequently taken by an unrelated ADR on 2026-07-07.

### Intent

Four open tickets describe a multi-repo change to the authentication boundary, and their stated source of truth does not exist. Per the org's own governance rule — *"No source of truth means the work is not ready"* — this epic cannot be started as written. Someone picking up #589 or #590 today would follow a citation to a Grafana dashboard ADR.

This is a decision ticket, not an implementation one: the outcome is either a real ADR or a closed epic.

### Source of truth (links)

- Epic #588, tickets #589, #590, #591
- `ai-helm-values`#300 (closed 2026-09-01 — reframed away by the owner on 2026-07-07, so it is **not** evidence this epic is live)
- `docs/adr/README.md` — no GitLab/multi-forge entry
- Governance doctrine: https://adorsys-gis.github.io/ai-governance/
- Found by the 2026-09-01 cross-repo open-issue triage sweep

### Current Behavior

#588/#589/#590/#591 are open and cite ADR-0077/ADR-0078. Those numbers resolve to dashboard and span-attribution ADRs. `lightbridge-repo-auth`#3 and #4 (GitLab support) are also open and blocked behind this epic; that repo has **zero** case-insensitive "gitlab" hits on `main`, and its migrations stop at `0003_blocked.sql`, still keyed on GitHub's `repository_owner_id`.

### Expected Behavior

Either:
- **(a)** a real ADR exists under a fresh number recording the multi-forge decision, and #588–#591 are re-anchored to it; or
- **(b)** the decision is that we are not doing this now, and #588–#591 + `lightbridge-repo-auth`#3/#4 are closed as not-planned.

### Acceptance Criteria

1. An owner decision recorded on this ticket: resurrect the ADR, or drop the epic.
2. If resurrected: the drafted ADR-0078 content is recovered (check unmerged branches and PRs) and re-filed under a **fresh** number — not 0078, which is taken and superseded.
3. If resurrected: #588, #589, #590, #591 have their "Source of truth" sections corrected to the new number, and `lightbridge-repo-auth`#3/#4 are linked to it.
4. If dropped: those six issues are closed with the reason, so nobody re-derives the plan from a dangling citation.
5. Either way, no open issue in any repo cites ADR-0077 or ADR-0078 for a GitLab/multi-forge claim.

### Out of Scope

- Implementing GitLab support. This ticket decides whether the epic has a basis at all.
- The three duplicate ADR numbers — separate defect, tracked by #1073. (They overlap only in that ADR-0077 is *also* one of the duplicated numbers, which is why the citation is doubly ambiguous.)

### Technical Context

Recovering the draft is the cheap first step: `git log --all --diff-filter=A -- 'docs/adr/*gitlab*' 'docs/adr/*multi-forge*'` and a search of closed/unmerged PRs. If the draft is gone, option (b) is likely the honest answer — nobody has touched the GitLab surface in `lightbridge-repo-auth` since.

### Risks

- ⚠️ Leaving this unresolved is the actual risk. Four tickets that look ready and cite an ADR by number will eventually be picked up by someone who does not check that the ADR exists — and this touches the auth boundary.
- If the epic is dropped, say so explicitly. A silently abandoned epic gets re-proposed in six months.

### Test Plan

Grep every open issue across `ai-helm`, `ai-helm-values`, `lightbridge-repo-auth` and `lightbridge-opencode-toolbeit` for `ADR-0077`/`ADR-0078` in a GitLab context; expect zero hits after this closes.

### AI Usage Declaration

Investigation and drafting by Claude Opus 5, during a cross-repo open-issue triage sweep on 2026-09-01. Every claim above was verified against `origin/main` — the ADR filenames, the absence of any GitLab ADR, and the citation text in each of the four tickets. AI output is not truth; the owner makes the call this ticket asks for.