What Problem This Solves

Maintainer-requested test performance work. Several unit and Control UI tests spent seconds loading unrelated runtime graphs, advancing every intermediate fake-clock timer, or sleeping for fixed delays even though the behavior had a precise completion signal.

Why This Change Was Made

The tests now isolate the dependency actually under test, keep the real process-level rejection-containment proof while removing its redundant negative child process, shorten a real timer schedule without weakening its reset requirement, fast-forward long fake-clock gaps, and replace fixed browser sleeps with causal request, state, bridge, mutation, or rendering barriers. CI configuration and production behavior are unchanged.

User Impact

No user-visible behavior change. Developers and CI get faster test feedback with the same behavioral and security coverage.

Evidence

• Fresh exact-head Blacksmith Testbox: 45 focused tests passed across unit-fast, infra, agents, and UI E2E shards in 1m29s.
• Before/after examples from the measured slow-test pass: auth-profiles.doctor.test.ts 54,344ms -> 163ms; exec.no-output-timer.test.ts 2,261ms -> 1,027ms; callback fatal proof file 4,579ms -> 3,945ms on the fresh box (1,126ms warm).
• Final targeted gates: oxfmt clean on all 10 files, core test types clean, oxlint 0 warnings/errors.
• Repo-native check:changed completed successfully before the final synchronization refinements; every final refinement then received focused remote proof and the final targeted gates above.
• Fresh branch-mode autoreview against origin/main: clean, no actionable findings.
• AI-assisted: yes. I understand and reviewed the resulting test behavior.