When Should a Test Stop Protecting the Past?

A test suite can preserve hard-won safeguards—and preserve assumptions the product has since outgrown. Keeping every old test may make useful changes harder; removing one too quickly can erase a protection people still rely on.

When product behavior changes, what evidence should guide whether a test is kept, revised, or retired? Who should be involved in that decision, and what would you review afterward to catch a safeguard that was lost—or a test that no longer serves users?