Startup‑time regression: deciding if a Hermes rewrite is worth it

The situation revolves around a React-Native team that recently upgraded their JavaScript engine from JSC to Hermes on version 0.71 of the framework. Reports indicate that user-reported complaints regarding the app taking too long to open have increased by 15% since the upgrade. The crash-free users metric has dropped from 99.6% to 98.9%, and the support team is handling approximately 30 extra tickets per day due to this issue.

Hermes improves startup time by pre-compiling JavaScript to bytecode and reducing the runtime's memory footprint. However, it only shows benefits when the bundle size is small enough to parse quickly and native modules do not block the main thread during initialization. Common regressions following a switch to Hermes include a cold-start time increase of 200-400ms, a large bundle size of 8MB where the bytecode loader has to read the entire file from storage, and native modules performing heavy work on the UI thread during initialization.

To determine if the Hermes migration is the cause of the slow startup and if a code rewrite can recover the lost start-up time, a systematic approach is recommended. This approach includes verifying the build uses Hermes, installing the APK on a clean device, warming up the device cache, capturing baseline cold-start metrics, enabling Hermes-specific performance logging, measuring native-module blocking, and comparing the results against a JSC baseline.

If the total time is significantly higher than the JSC baseline and the Hermes bytecode load exceeds certain thresholds, it may indicate that a rewrite is necessary to recover the lost start-up time before investing weeks in refactoring.