Sketching blast radius before you rename a service

Renaming feels trivial until midnight pages prove otherwise. We start by drawing dependency cones on butcher paper, then annotate synchronous versus asynchronous callers with different ink weights. That contrast alone has prevented three bad cutovers in recent cohorts. The second paragraph is about measurement: we ask teams to paste traceroute summaries and queue depth screenshots beside the diagram. Nothing fancy—just enough signal for the group to argue from evidence instead of memory. Finally, we capture the conversation in a decision record stub before anyone opens an editor. That ordering matters: words before code reduces the odds of a heroic revert. If you try this, expect the first session to feel slow; speed shows up on the fifth rename when the ritual is muscle memory.