ci(diagnose): retire the 1874 diagnose loop

[thymikee]

Summary

Follow-up to the review on #2845. The review noted that 1874-diagnose.yml defaults to testBareTypeUsesTappedInputWhenSoftwareKeyboardIsHidden. After #2845 that test asserts route xctest-element and never reaches a [DEBUG-1874] commit wait. Of the two options (repoint the default or retire the workflow), this PR retires it.

• The loop never reproduced ios: fill's synthesized-replacement route lands a corrupted value (traced: 11 chars requested, 7 observed, frozen) #2080. The default test is a bare type. ios: fill's synthesized-replacement route lands a corrupted value (traced: 11 chars requested, 7 observed, frozen) #2080's synthesized-first-responder-replacement route fires only for fill on a penalized XCTest channel, so the default test did not reach it before refactor(ios-runner): restrict the tap-witness synthesized type route to the bare submit key #2845 either.
• No existing test can be the new default. Only testTypeTextReliablyPacesSynthesizedReplacementThroughProductionCaller reaches the replacement commit wait. It uses a recording synthesizer against Springboard, so every poll reads nil and it always ends notObserved at the 3 s deadline. The cadence it logs is fixed and synthetic.
• A new test would not help either. A penalized coordinate-fill test could reach the route, but cadence records lengths only (fix(ios): report an unobserved text commit instead of a partial success #1924). ios: fill's synthesized-replacement route lands a corrupted value (traced: 11 chars requested, 7 observed, frozen) #2080 needs the corrupted value, and its triage asks for a local reproduction with test-only instrumentation.
• The only other use is closed. A default dispatch still flagged slow bare-type passes by their type-all duration, which only mattered for ios: text-entry lanes fail intermittently on unrelated branches (unpinned simulator keyboard state) #1874, now closed.

Deleted: the workflow, scripts/diagnose-1874-iteration.ts, its test, and that test's vitest.config.ts entry (4 files, −370). The runner's [DEBUG-1874] lines stay. The #2080 trace came from them in the smoke lane's runner.log, and existing tests pin their format and redaction.

Validation

• pnpm check:affected --run passed: 59 checks selected (including fallow and build), GitHub-authoritative lanes not run locally. It ran at 120233a7a, which has the same tree as the pushed f6888782f; only the commit message was reworded.
• pnpm check:xctest-selection: 308 methods, 0 reachable by no lane.
• git grep finds no leftover references.
• An adversarial review found no blocking issues. Its two wording corrections are applied above.

🤖 Generated with Claude Code

[github-actions]

Size Report

Metric	Base	Current	Diff
Installed (including dependencies)	4.80 MB	4.80 MB	0 B
Package (unpacked)	4.80 MB	4.80 MB	0 B
Package (download)	1.44 MB	1.44 MB	-2 B

Startup median (7 runs, lower is better):

Scenario	Base	Current	Diff
CLI --version	27.7 ms	27.7 ms	-0.0 ms
CLI --help	82.5 ms	80.7 ms	-1.8 ms

[thymikee]

Reviewed at f688878. The deletion looks clean: it removes the diagnose workflow, the script, its test, and the matching line in vitest.config.ts. It touches no production code.

The failing Smoke Tests job is a live iOS run. It fails at get text id="automation-event-name" because the selector did not match. This diff does not touch that path, so the failure is likely unrelated. I did not check whether it also fails on main, so a re-run or a separate look is worth it.

[github-actions]

PR Preview Action v1.8.1
Preview removed because the pull request was closed.
2026-09-24 10:44 UTC