fix(notifications): rescue notification_deliveries rows stranded at pending by a failed enqueue

[bitfathers94]

fix(notifications): rescue notification_deliveries rows stranded at pending by a failed enqueue

evaluateAndEnqueueNotificationDeliveries commits each fresh delivery row
before enqueuing its notify-deliver job, so a rejected JOBS.send leaves the
committed rows at pending. A client retry finds them already present
(idempotent) and omits them, so no deliver job is ever re-sent and the
notification stays invisible to the recipient until the 90-day retention
sweep deletes it.

Add a periodic sweep, mirroring sweepStrandedPendingClosures: a bounded
lookback window, a grace period so an in-flight row is not mistaken for a
lost one, and a minimum re-sweep interval so a row stuck for another reason
is retried periodically rather than every tick. A new repository query lists
pending rows within the window, the sweep re-enqueues notify-deliver for each
and fails open on a DB read error, and it is wired into the periodic queue
tick alongside the sibling repair scans.

Closes #9320

Validation

Verified locally on this branch before opening:

• npm run typecheck
• npx turbo run build:tsc build:verify
• npm run test:coverage — patch coverage 100.0% of changed lines

[superagent-security]

Superagent didn't find any vulnerabilities or security issues in this PR.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 76.15%. Comparing base (995cd78) to head (ea1f546).
⚠️ Report is 4 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #9393      +/-   ##
==========================================
+ Coverage   75.38%   76.15%   +0.77%     
==========================================
  Files         275      278       +3     
  Lines       58023    60168    +2145     
  Branches     6181     6882     +701     
==========================================
+ Hits        43739    45821    +2082     
- Misses      14014    14023       +9     
- Partials      270      324      +54     

Flag	Coverage Δ
backend	97.06% <100.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
src/db/repositories.ts	96.79% <100.00%> (ø)
src/notifications/stranded-delivery-sweep.ts	100.00% <100.00%> (ø)
src/queue/job-dispatch.ts	99.46% <100.00%> (ø)

[loopover-orb]

Tip

✅ LoopOver review result - approve/merge recommended

_{Review updated: 2026-07-27 16:08:10 UTC}

5 files · 1 AI reviewer · no blockers · readiness 98/100 · CI green · clean

✅ Suggested Action - Approve/Merge

• safe to merge

Review summary
Adds a periodic sweep (`sweepStrandedNotificationDeliveries`) that rescues `notification_deliveries` rows stuck at `pending` when the accompanying `notify-deliver` enqueue was lost, mirroring the existing `sweepStrandedPendingClosures` pattern almost line-for-line (grace window, lookback bound, per-row audit-event spacing gate, fail-open on DB read errors, bounded scan limit). The new repository query correctly bounds both ends of the `createdAt` window and orders oldest-first under the limit, and it's wired into the `agent-regate-sweep` tick alongside its sibling repair scans with the same best-effort `.catch(() => null)` pattern so a failure here can't cost the tick its re-gate work. Tests seed rows directly via raw SQL to control `created_at`/`status` and cover grace/lookback edges, the already-delivered no-op case, DB-read failure, unreachable-JOBS-send failure, and the audit-ledger-read/-write failure arms — this is solid, well-targeted coverage of the real code path.

Nits — 4 non-blocking

• The spacing gate does one extra `countRecentAuditEventsForActorAndTarget` DB round-trip per stranded row (up to 500/tick) — bounded and consistent with the sibling watchdog, but worth noting as the tick's main added DB cost.
• Comment in stranded-delivery-sweep.ts:13 and repositories.ts:2319 references the '90-day retention sweep' only in prose, not as a shared named constant — fine for now but drifts silently if that retention window ever changes elsewhere.
• The external brief's 'debug leftover console' flags on job-dispatch.ts:321 and stranded-delivery-sweep.ts:54 are false positives — both are the same structured `console.log`/`console.warn` JSON logging convention used by every sibling reconciler in this file, not debug leftovers.
• Consider whether `STRANDED_NOTIFICATION_REQUEUED_EVENT` audit rows should eventually be pruned by the retention sweep like the deliveries themselves, to avoid unbounded audit_events growth for chronically-stuck rows (same question applies to the pending-closure sibling, so likely already handled/acceptable).

Decision drivers

• ✅ Code review — No blockers (1 reviewer)
• ✅ Gate result — Passing (No configured blocker found.)

Context & advisory signals — never blocks the verdict

Signal	Result	Evidence
Linked issue	✅ Linked	#9320
Related work	✅ No active overlap found	No same-issue or scoped active PR overlap found.
Change scope	✅ 20/20	Low review scope from cached public metadata (1 linked issue).
Validation posture	✅ 25/25	PR body includes validation/test evidence.
Contributor workload	✅ 10/10	Author activity: 80 registered-repo PR(s), 54 merged, 0 issue(s).
Contributor context	✅ Confirmed Gittensor contributor	bitfathers94; Gittensor profile; 80 PR(s), 0 issue(s).
Improvement	✅ Minor	risk: clean · value: minor · LLM: moderate

Linked issue satisfaction

Addressed
The PR adds a DB query for stale pending notification_deliveries, a sweep function mirroring sweepStrandedPendingClosures with grace period, lookback window, and re-sweep spacing (via audit-event check), fails open on DB read errors, filters by status:"pending" so already-delivered rows aren't re-enqueued, and wires the sweep into the periodic queue tick alongside the sibling reconcilers with the

Review context

• Author: bitfathers94
• Role context: outside_contributor
• Public audience mode: oss maintainer
• Lane context: Repository is configured for direct PR review.
• Public profile languages: TypeScript
• Official Gittensor activity: 80 PR(s), 0 issue(s).
• PR-specific overlap: none found.

Contributor next steps

• Keep the PR focused and include validation evidence before maintainer review.

Signal definitions

• Related work = same linked issue, overlapping active PRs, or title/path similarity.
• Change scope = cached public metadata such as size labels, draft state, and review-burden hints.
• Validation posture = whether the PR provides enough public validation/test evidence for maintainer review.
• Contributor workload = public contributor activity and cleanup pressure, not a repo-wide quality failure.
• Contributor context = public GitHub/Gittensor identity context; non-Gittensor status is not a blocker.

🧪 Chat with LoopOver

Ask LoopOver a question about this PR directly in a comment — grounded only in the same cached, public-safe facts shown above, never a new claim.

• @loopover ask <question> answers contribution-quality Q&A with source citations and freshness.
• @loopover chat <question> answers in natural prose from cached decision-pack facts via local inference (maintainer/collaborator; read-only).
• A plain-language @loopover mention with a real question is routed to the closest matching read-only command automatically — no exact syntax required.

Full command reference: https://loopover.ai/docs/loopover-commands

🧪 Experimental — new and may change.

_{🟩 Safe / merged · 🟦 Advisory · 🟨 Held for review · 🟥 Blocked / closed}

---

💰 Earn for open-source contributions like this. Gittensor lets GitHub contributors earn for the work they already do — register to start earning →.

Checked by LoopOver, a quiet PR intelligence layer for OSS maintainers.

• Re-run LoopOver review

[loopover-orb]

LoopOver approves — the gate is satisfied and CI is green.