[taras]

Why

L2 needs a real unanswered question before it checks the element's lifecycle. Component import can still be in enter after entry admission, so a finite scheduler poll does not establish question readiness.

What changes

Before: L2 polls for up to 1,000 scheduler turns until some lifecycle occurrence reports a wait.

After: L2 subscribes to question changes before checking the pending question, then asserts the entire Elicit:active:question reading. Answering the real question still leads through joined entry teardown to an empty reading and non-live session.

How to verify it

The added regression holds Elicit import and checks the entire Elicit:enter: reading with no pending question. After release, it obtains the real question, checks the full lifecycle and waiting-element assertions, answers and joins completion. A missing question wait still fails the assertion after question readiness completes.

Retained focused runs passed on the candidate verified to match signed head 06e31348460a0320d9db55dde993c3511ca24848:

• deno task test packages/cli/tests/repl-lifecycle.test.ts — 1 passed, 9 steps, 0 failed.
• deno task test packages/cli/tests/repl-execution.test.ts packages/cli/tests/repl-lifecycle.test.ts — 8 passed, 42 steps, 0 failed.

Scope

Included

• Question readiness and one controlled delayed-import regression in packages/cli/tests/repl-lifecycle.test.ts.

Intentionally unchanged

• Production lifecycle reporting, provider ownership, answer validation, persistence and replay.

New abstractions

• A private scoped nextQuestion helper serves ordinary L2 and the delayed-import regression. Its subscription ends when the read returns, fails or is cancelled.
• Each new abstraction has multiple concrete uses or a clear justification.
• No speculative functionality is included.

Risks and limitations

The controlled delay demonstrates the pre-question interval. The exact scheduling of the historical failures was not reproduced.

Scope confirmation

• Every changed file supports the purpose described above.
• Unrelated cleanup and formatting changes are excluded.
• No generated or mechanical files changed.
• The description matches the final diff and retained test results.

Refs #935.

[github-actions]

PR #981: 🧪 Wait for pending questions in REPL lifecycle tests

1 files, +73 / -24

Scope

✅ PR scope looks good.

Structural

✅ No structural bloat detected.

Slop

✅ Slop indicators look low.

Static Analysis

✅ Oxlint found no issues.

Correctness

FILE: packages/cli/tests/repl-lifecycle.test.ts
PATTERN: Speculative abstractions
CONCERN: nextQuestion function has only 1 consumer (this test) but implements general-purpose logic
QUESTION: Would this logic benefit from being in a shared test utility or eliminated in favor of direct assertions?

FILE: packages/cli/tests/repl-lifecycle.test.ts
PATTERN: Speculative abstractions
CONCERN: nextQuestion implementation includes custom iterator handling that may not justify standalone abstraction
QUESTION: Could this logic be inlined to reduce cognitive overhead without sacrificing readability?