test(ios): speed up fixtures and run native regressions nightly

Merged
TH
8h ago
+657-309

Summary

  • Run the full native iOS regression suite nightly. PRs keep 338 host tests, four generated iOS branch checks, and public CLI E2E smoke.
  • Reuse native fixtures and check existing fixtures before XCTest's one-second polling wait. Retain native fault/identity oracles and remove the duplicate drag row.
  • Rebase onto #3820's readiness fix and preserve its production implementation. Scope expands to CI scheduling and profiling evidence: 26 files, 966 gross lines. Runtime +3/-2 (comments), tests +214/-234, CI/tooling +31/-73, evidence +409/-0.

Validation

  • Tested 14f32fb218ac9e5608012bf6a5805652dc61f7fc: affected gate passed, exactCommit=true, checkoutChanged=false (36 passed, 25 skipped, two deduplicated).
  • Same 118 methods: 718.284s → 484.579s local wall time. This includes upstream changes and warmer simulator state. Per-method results and attribution.
  • Those 118 plus 331 complementary methods cover all 449 nightly identifiers across runs: 448 passed, one expected failure, no unexpected failures or skips. Native source SHAs: 19573d42b0 / 7bb347f771.
  • Xcode 27.1: 15 focused checks passed; tooling: 75 passed. Five production mutants detected; restored controls pass. Four-method PR lane passes in 4.4s locally, including its count gate. Independent review found no issues.
  • CI at 14f32fb: Coverage and Integration passed. Android is blocked by #3831, a capture-stall signature also observed on main; cause unresolved, no infrastructure retry. Apple passed in 23m49s; iOS platform checks took 2m55s (four tests, 0.8s bodies). Native fault regressions now run after merge nightly.