Improves the disabled-store health test so a busy executor scheduler no longer causes a false "degraded" diagnosis; the test now retries the health read after contention clears before judging the storeable. This also adds a benchmark descriptor for the retry behavior and lets the test accept extra contended reads without hanging.
Reuses the scheduler contention detail string through a shared constant, so tests compare against one source of truth.
Adds a helper that retries up to 200 health reads, yielding between attempts, and asserts at least four reads since the scheduler thread can also hold the lock.
Adds a helper that deterministically holds the scheduler lock while a report is built; a panic drops the release sender so the holder never blocks forever.
Adds the land-health-contention-retry.json harness descriptor.
Confidence Score: 5/5
The PR appears safe to merge; both earlier test issues are fixed.
Summary
The PR makes the disabled-store health test retry when the scheduler is busy.
The disabled-store health test waits for scheduler contention to clear.
