When product deadlines are tight, how do you decide which technical debt to tackle now versus defer? What is one decision your team made about debt that proved right in hindsight, and what simple rule guided that call?