ci: ci-status fails at its 540 s wait while the full run is still running

[kyle-sexton]

ci-status waits up to 540 s for a successful ci-lanes status. When the lane run takes longer, or a duplicate run is still in flight, ci-status fails with "no successful ci-lanes status ... re-run the full workflow", and the PR stays red until someone re-runs it by hand. The re-run then passes.

Seen during the 2026-10-01 drain:

• fix(planning): resolve loop.md spoke paths without the plugin-root token #5688: twice, once on each head (both re-runs passed).
• chore: remove docs/specs and the topic-docs convention #5719: twice.
• feat(planning): re-ground interview recommendations and apply the digested skill text #5761: once.
• feat(source-control): resolve promotion evidence through the bootstrap in the babysit-loop cycle #5677: once, while a duplicate run was still in flight.

Each costs a manual re-run and 10-40 minutes per PR, and an unattended merge lane reads the red check as a failure.

Acceptance:

• ci-status either waits for the lane run it depends on to finish, or re-checks after the lanes complete, so a slow but green run never leaves a red ci-status behind.
• A test or a probe PR shows a lane run longer than 540 s ending with a green ci-status and no manual re-run.

Related: #4670 (stale failure during a re-run).