PR 6635 Review Follow-ups Implementation Plan

Goal

Close the correctness gaps in multi-workspace daemon-managed channel workers without adding workspace-qualified channel control APIs.

Implementation

1. Use one cwd resolver for parsing and ownership grouping. Resolve ordinary relative channel cwd values from the settings workspace and preserve absolute and home-relative behavior.
2. Resolve and validate the complete channel-to-workspace plan at listener startup. Reject unknown, ambiguous, untrusted, or uncanonicalizable owners before exposing a usable daemon handle, then freeze that plan for runtime creation.
3. Make the worker group the lifecycle boundary. Initial startup rolls back on partial failure; daemon-wide reload coalesces concurrent requests and stops the fleet if any restart fails; shutdown waits for reload completion.
4. Publish a newly built runtime app only after all workers are ready. Keep health degraded on startup failure.
5. Preserve single-workspace wire shapes while adding validated per-workspace pidfile, daemon-status, diagnostic, and SDK fields for multi-workspace mode.
6. Cover relative cwd semantics, canonicalization failures, trust validation, partial startup/reload failures, two-workspace orchestration, pid cleanup, status diagnostics, and compatibility behavior with focused tests.

Verification

Run affected CLI and SDK tests, formatting and lint checks, then repository build and typecheck. Exercise a two-workspace daemon-managed channel scenario and audit the final diff for lifecycle races, compatibility regressions, and unnecessary API surface.