Rehearse the merge in a throwaway worktree before choosing a hotfix's base branch
Before targeting a hotfix at a production branch, actually perform the merge the release automation will perform, instead of reasoning about whether it will be clean: git worktree add --detach /tmp/probe origin/<target> then git merge --no-commit --no-ff <your-sha> gives you the real conflict list in seconds and costs nothing, because you delete the worktree afterwards. This matters most when the production branch and the integration branch have structurally diverged — files that exist on only one side, or a refactor that moved helpers — since any edit to the refactored regions conflicts regardless of how small the fix is. A conflicted forward-merge typically fails the reconciliation job without pushing, which holds the release freeze and forces hand-merging under pressure, so discovering it up front can flip the decision to targeting the integration branch and letting normal promotion carry it. Two things made the call concrete: measuring exposure from real data rather than intuition, and noticing that a client or frontend in a separate repo deploys independently of the frozen backend, so normalizing input there closes the practical gap while the real fix takes the slow safe route. Also rehearse a careless resolution by taking one side wholesale and running the incoming tests — if they fail loudly, a bad merge cannot land quietly.