How a Two-Stage Merge Strategy Prevented Silent Production Code Regression

A development team discovered that their production server was running a PHP file that had been manually hot-patched months earlier but never committed back to the git repository. The drift was caught by chance when a developer noticed unfamiliar code while preparing a bug fix, and git blame revealed the lines had no history. Had the fix been deployed from the outdated local repo, a half-year-old internationalisation feature handling English and Japanese display logic would have been silently overwritten. The team resolved this with a two-stage strategy: first syncing the production file into the local repo as a standalone commit, then adding new functionality on top without altering existing code. MD5 hashing of preserved code sections was used before and after deployment to verify no production behaviour had changed.