How a shared Chrome instance silently broke SEO for a 160-post React SPA

A development team running react-snap to pre-render roughly 168 routes for their React single-page app discovered that sharing system memory with an open Google Chrome browser caused the Puppeteer-driven crawl process to be killed partway through, typically around routes 45–59. Because the preceding webpack build step still exited successfully, the CI/CD pipeline reported a clean build even though two-thirds of blog posts and the homepage were deployed as empty client-side-rendered shells. Search engines and ad reviewers flagged the thin content before any internal monitoring caught the problem, with indexing quality dropping as the first visible signal. The root cause was pointing react-snap at a system-installed Chrome rather than a bundled browser, making available memory dependent on what else was running on the machine. The team now recommends fully quitting Chrome before building, verifying the crawl count matches total routes, and adding a content-assertion check as a hard deployment gate.