A scraped product with nil price_per_round_cents made Prices.create_snapshot
return {:error, changeset}, which blew up the {:ok, _} match in the runner,
crashed the whole perform, and after 3 retries discarded the job. Nothing
reseeds the chain once the job is discarded, so prod scrapes silently died
on Mar 17 and stayed dead until the app was restarted.
- Runner handles snapshot insert errors by logging and continuing instead
of raising.
- ScrapeJob wraps each Runner.run in try/rescue so one retailer crashing
doesn't take down the whole cycle.
- ScrapeJob moves schedule_next into an after block so the chain is always
reseeded, even if the scrape loop crashes.
Graham McIntire
2efb35dc48