Problem
v0.9.0 should not only add WhaleFlow, Model Lab, and new UI surfaces. It also needs a stabilization gate so old correctness and usability bugs do not ship under a larger feature release.
Scope
Make a release-blocking stabilization lane that triages and closes or explicitly defers the highest-risk runtime issues:
- Windows input and rendering:
- Large repository startup and context loading:
- Sub-agent reliability:
- Long-running command/live-state correctness:
- Prompt/UI mode clarity:
Acceptance criteria
- Each listed issue has a disposition: fixed, duplicate, not repro, deferred with reason, or still release-blocking.
- At least one focused regression test or smoke script exists for each fixed blocker class.
- Windows fixes are verified on Windows or documented as unverified with a precise manual test plan.
- Large-repo startup is tested against a synthetic many-file workspace.
- Sub-agent timeout behavior is tested without requiring real provider secrets.
- Release notes call out any remaining known risks.
Non-goals
- Do not fold unrelated v0.9.0 feature work into this lane.
- Do not mask failures by relaxing tests or hiding errors.
Problem
v0.9.0 should not only add WhaleFlow, Model Lab, and new UI surfaces. It also needs a stabilization gate so old correctness and usability bugs do not ship under a larger feature release.
Scope
Make a release-blocking stabilization lane that triages and closes or explicitly defers the highest-risk runtime issues:
Acceptance criteria
Non-goals