[Bug] Mirror never sets `is_transferred=true`

[Martin-Blau]

Summary

Shortly, gittensor's transferred-issue anti-gaming gate is dead on the mirror path.
das-github-mirror declares issues.is_transferred BOOLEAN NOT NULL DEFAULT FALSE (packages/db/03_issues.sql:18) and exposes the column on the public miner API (packages/das/src/api/miners/miners.service.ts:83,149 — both getIssues and the inlined linked-issue rows on getPullRequests). No code path in das-github-mirror ever writes is_transferred = true:

• webhook/handlers/issue.handler.ts does not branch on payload.action === "transferred". It upserts the same set of fields for every action; isTransferred is omitted from the upsert payload, so the DB default (false) sticks.
• webhook/github-fetcher.service.ts:869-892 backfill issue query only requests timelineItems(itemTypes: [LABELED_EVENT, UNLABELED_EVENT], first: 30) — TRANSFERRED_EVENT is not in the list.
• Grep isTransferred|is_transferred across packages/das/src shows zero write sites.

On the gittensor side, the mirror path treats this field as the first anti-gaming gate in mirror_scan.py:_classify_issue:

if issue.is_transferred:
    bt.logging.debug(f'  issue #{issue.issue_number} ({issue.repo_full_name}): ignore (transferred)')
    return 'ignore'

Because the mirror always returns is_transferred=false, this branch is unreachable in production. The mirror-path docstring (mirror_scan.py:12-19) explicitly lists not issue.is_transferred among the gates, but the gate never fires.

Why this is a regression and not a fresh oversight

The legacy (pre-mirror) scan path used to detect transfers and had this exact class of bug — fixed by issues #404 and #405 with PRs #415 / #420 / #429. The mirror migration shipped in PR #796 routed OSS scoring + issue discovery through the mirror service for mirror_enabled repos. The mirror integration carried over the field shape (is_transferred returned in API responses, parsed into MirrorIssue.is_transferred, gated in _classify_issue) but not the producer — nothing on the mirror side actually populates it.

Net result: the anti-gaming protection added by #420/#429 silently regressed for every repo on the mirror path.

Reproduction

1. Pick a mirror-enabled tracked repo. Open an issue, then transfer it to a different repo using GitHub's UI.
2. GitHub fires issues webhook with action="transferred". The mirror's IssueHandler.handle runs and upserts the row — is_transferred is never touched.
3. Confirm directly:

   SELECT issue_number, state, is_transferred
     FROM issues
    WHERE repo_full_name = '<owner>/<repo>'
      AND issue_number   = <n>;

   is_transferred is false.
4. GET /api/v1/miners/<authorId>/issues returns the row with "is_transferred": false.
5. In mirror_scan._classify_issue the issue does not short-circuit at line 431; it falls through to normal 'solved' / 'not-solved-closed' classification.

Expected behavior

Mirror must populate is_transferred=true for any issue that has been transferred. Two producer paths are needed for parity with the rest of the schema:

1. Live webhook (issue.handler.ts): when payload.action === "transferred", set data.isTransferred = true in the upsert. (Optional: also persist the destination from payload.changes.new_repository / payload.changes.new_issue for observability — not required to fix the gate.)

2. Backfill (github-fetcher.service.ts:869): add TRANSFERRED_EVENT to timelineItems(itemTypes: [...], first: 30) and, when seen, set isTransferred=true on the upsert. Without this, transfers that happened before App install or before the most-recent backfill window stay invisible.

Impact

• Anti-gaming gate documented at mirror_scan.py:18 (not issue.is_transferred) is dead. Transferred issues classify as 'solved' or 'not-solved-closed' and feed solved_count / valid_solved_count / closed_count, distorting issue-discovery credibility on every mirror-enabled repo where transfers occur.
• Removes one of the surfaces previously hardened by [BUG] Transferred issues bypass anti-gaming credibility penalty in issue discovery scoring layer(score.py) #404 / [BUG] Transferred issues are never detected in repo-centric scan #405 / fix: handle transferred issues in issue discovery scoring #420 / fix: bypass credibility penalty for transferred issues #429. Same gaming vectors those PRs closed are reopened on the mirror path.

Notes

• Tests covering the gate exist (tests/validator/issue_discovery/test_mirror_scan.py:184 test_transferred_ignored, tests/validator/oss_contributions/mirror/test_scoring.py:549 test_transferred_blocks) — they pass because they construct is_transferred=True directly. They don't exercise the producer side, which is where the real bug lives.
• This issue is purely about the producer (mirror). The gittensor consumer is already correct.