Work out why a Kipory flow behaved the way it did — find the run, read its never-sampled step log, see what it wrote, open its trace to learn what each output slot actually held, and compare a good run with a bad one. Use when a flow returns the wrong thing, an endpoint 502s, a scheduled run failed, a record is stuck, output changed and nobody knows which step moved, or the user asks "what happened". Not for fixing the flow (that is build) and not for asserting it stays fixed (that is prove).
