Fix flaky unit tests by replacing fixed sleeps with vi.waitFor

[umair-ably]

Summary

Four unit test files used fixed setTimeout waits to synchronise with async work inside runCommand. When the work took longer than the hard-coded delay, the assertion ran against stale state and the test failed intermittently.

Replaced those fixed sleeps with vi.waitFor(() => expect(...)), which polls the condition and eliminates the race.

Files changed:

• test/unit/commands/logs/connection-lifecycle/subscribe.test.ts — 5 tests scheduled setTimeout(..., 50) to fire the message callback, but channel.subscribe() wasn't always captured within 50 ms, so the callback check saw null and silently dropped the message. (This was the actual failure that triggered the investigation.)
• test/unit/commands/interactive-terminal-behavior.test.ts — 2 tests slept 100 ms after rl.emit("line", ...) then asserted on async prompt/close state.
• test/unit/commands/spaces/locations/subscribe.test.ts — 2 tests slept 50 ms after starting the command before invoking the captured subscribe handler.
• test/unit/hooks/interactive-did-you-mean.test.ts — slept 30 ms after the hook returned before asserting on readline state restoration.

Test plan

• pnpm exec eslint . — 0 errors (7 pre-existing warnings)
• pnpm test:unit — 2292 passed, 0 failed
• Targeted re-run of the four modified files — 42 passed

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
cli-web-cli	Ready	Preview, Comment	Apr 20, 2026 7:13pm

[claude-code-ably-assistant]

Walkthrough

Four unit test files used fixed setTimeout delays to synchronize with async work inside runCommand. When the async work exceeded the hard-coded delay, assertions ran against stale state, causing intermittent (flaky) failures. This PR replaces all fixed sleeps with vi.waitFor(() => expect(...)), which polls until the condition holds and removes the timing race entirely.

Changes

Area	Files	Summary
Tests	test/unit/commands/logs/connection-lifecycle/subscribe.test.ts	5 tests: replaced setTimeout(..., 50) callbacks with vi.waitFor; the old delay wasn't enough to guarantee channel.subscribe() was captured before the message callback fired
Tests	test/unit/commands/interactive-terminal-behavior.test.ts	2 tests: replaced 100 ms post-rl.emit sleeps with vi.waitFor assertions on async prompt/close state
Tests	test/unit/commands/spaces/locations/subscribe.test.ts	2 tests: replaced 50 ms pre-handler-invocation sleep with vi.waitFor
Tests	test/unit/hooks/interactive-did-you-mean.test.ts	1 test: replaced 30 ms post-hook sleep with vi.waitFor on readline state restoration

Review Notes

• No production code changes — all 4 modified files are test files; zero risk to runtime behavior.
• Net deletion (+127 / −132): the vi.waitFor form is marginally more verbose in a few spots but eliminates the arbitrary timeout values entirely.
• No new dependencies — vi.waitFor is already available from Vitest.
• Flakiness root cause confirmed: the logs/connection-lifecycle/subscribe tests were the reported failure; the other three files were found during the same investigation and fixed proactively.
• No migration, deployment, or breaking-change considerations.