How We Fixed a Critical Magento 2 Checkout Bug in Under 2 Hours

The Emergency Call

At 2:15 PM on a Friday, we received an emergency request through our Emergency Rescue Platform. A high-volume Magento 2 retailer's checkout had completely stopped working. Customers could add items to the cart, but clicking 'Proceed to Checkout' resulted in an endless spinner and eventually a 504 Gateway Timeout.

Business Impact: They were losing approximately $15,000 per hour in abandoned carts.

The Investigation

Our senior Magento engineers immediately jumped into the AWS environment. We bypassed the Varnish cache to replicate the issue directly on the origin NGINX nodes. The PHP-FPM error logs showed nothing, but New Relic highlighted a massive spike in database query time specifically on the `sales_quote` tables.

The Root Cause

A third-party reward points extension, which the client's previous agency had installed the night before, was running an unindexed table scan on every checkout initialization to calculate potential points. Under low traffic (testing), this was unnoticeable. Under Black Friday levels of traffic, it caused complete database lock contention.

The Fix

• Minute 15: We safely disabled the offending reward points module via CLI and cleared the compiled code.
• Minute 25: Checkout immediately stabilized. We confirmed with the client that processing orders was more important than issuing reward points temporarily.
• Minute 90: We patched the module by adding the missing foreign keys and indexes to the database schema, and rewriting the query to prevent full table scans.
• Minute 110: Deployed the patched module to staging, verified, and pushed to production.