The question

Why is “this code looks suspicious” a weaker starting point than an observed lifecycle anomaly plus an ownership model?

Working model

Begin with the resource or state: who created it, which lifetime owns it, what transfers or ends that ownership, what completion means, and what happens on success, error, cancellation, restart, warm state, and cold state.

Then require a target-native regression that fails on a clean baseline before promoting a candidate. A plausible source theory is a hypothesis; a discriminator makes it a finding. Invalid probes are useful when you record how setup answered the wrong question.

Use one finding to inspect adjacent branches, but make each branch earn promotion independently.

Provenance

Synthesized from the Vite correctness-lane process notes at Fieldwork revision 49e8c7dabbc75f021b09d243b1d95ecac215bb07. These are process observations, not Vite requirements.

Apply the model

Pick one object from a running development server—a watcher, plugin container, dependency cache, child process, or bundle—and answer:

1. Who creates it?
2. What generation owns it?
3. What operation ends that generation?
4. What does settled cleanup mean?
5. Which warm/cold or success/failure pair could make the theory lose?

Only then sketch the smallest regression.