claude-code - 💡(How to fix) Fix Claude deployed VividSeats scraper that silently never worked — zero data collected since deployment [2 comments, 3 participants]

Severity: HIGH — Silent deployment failure, total data source loss

Date discovered: 2026-04-06 | Duration: weeks since original deployment

What happened

Claude deployed a VividSeats OC browser scraper. The scraper had a fundamental flaw: it timed out in Phase 1 on a React SPA selector that never resolved (VividSeats uses PerimeterX bot protection). The script never reached Phase 2 where data was actually collected.

Result: Zero VividSeats price snapshots were ever saved. The scraper appeared to run (cron fired, process started, no errors logged) but collected nothing.

This was only discovered during a full manual audit weeks later.

Impact

• Entire VividSeats data source missing from price comparison system since deployment
• Competitive pricing intelligence had a major blind spot for the entire period
• Required a complete rewrite to fix (switched to Hermes API via OC browser evaluate)
• User had been operating under the assumption VividSeats data was flowing

Root cause: No post-deploy verification

Claude added the cron, declared it deployed, and moved on. There was no verification that:

1. The scraper actually collected data
2. At least one row was written to the database
3. The log showed successful scrape completion

Pattern

This is the same failure mode as the Google Hotels NameError (separate issue) — Claude deploys scrapers without verifying they actually collect data.

Requested resolution

• Refund of session credits
• Fix: Any scraper deployment must include: run manually, verify DB has new rows, confirm log shows success. Only then is it "deployed."