# 2026-07-14 — Worker stopped after PnP push; parallel rotate

## Finding
- Last Ops ingest run: **R-74** (PnP) closed clean **2026-07-13 12:54:45**.
- Heartbeat frozen at push start — no pulse during long work, then worker never resumed.
- Ingest runs only open on **push**, so a dead worker leaves Scrape runs on old rows.
- Dokploy MCP/SSH may be needed to restart the worker container.

## Fixes in repo
1. Rotate: start next retailer after **45 min** if still busy; **max 2** parallel.
2. Heartbeat keepalive every **5 min** during scrape/push.
3. `abort-stale` CLI (39 stuck open runs cleaned 2026-07-13).

## Redeploy
Push, then rebuild/restart the Dokploy **worker** service so Ops gets new scrape runs.