Claude deployed Google Hotels scraper with NameError bug — 12 days of silent failure, 72,000 missing data points, unrecoverable

Severity: HIGH — Silent production failure, permanent data loss

Date discovered: 2026-04-06 | Duration: ~12 days

What happened

Claude deployed a Google Hotels Playwright scraper (playwright_google_scraper.py) with a NameError: DELAY_SECS on line 234. The variable was used in an error-handling branch but never defined.

The script crashed silently on every run:

• No log file was created (log path also had a bug)
• The cron showed as running but produced nothing
• Zero data was collected for 12 days
• Discovered only during a full audit of all scrapers

Impact

• 12 days × ~6,000 price points/day = ~72,000 missing hotel price data points
• Data is permanently unrecoverable — cannot backfill Google Hotels historical prices
• Revenue intelligence system had a 12-day gap in its primary data source
• No alert fired because the failure was silent

Root cause: No deployment verification

Claude deployed the scraper without:

1. Running it once manually to verify it executes without errors
2. Checking that the log file was being written
3. Verifying at least one successful data row in the database

Requested resolution

• Refund of session credits for the deploy session + the audit session required to discover it
• Fix: After deploying any scheduled scraper/cron job, Claude must verify: (1) it runs without errors, (2) it writes to its log, (3) at least one DB row is inserted. "Deployed" should mean "verified working," not just "file written and cron added."