JP Journal lab · deterministic model + bounded observation
Main-thread yield lab
Compare the same work as one blocking task, timer-separated chunks, and prioritized continuations. Then run a small local sample without turning one device into a platform claim.
Truth boundary: the deterministic simulator models declared job costs. The live sample measures only this browser and device. Neither proves normative event-loop ordering or universal performance.
Modeled queue timeline
Modeled results
| Strategy | Total | Worst input wait | Yields |
|---|
Live sample
| Selected path | — |
|---|---|
| Observed total | — |
| Chunks | — |
| Cancellation | — |
Non-normative: background state, JIT, power, thermal conditions, and other work can change this sample.