[JAORMX]

Summary

Run TestToolhiveProbeDeadlinePublishesHealthySibling inside testing/synctest, following the neighboring stale-refresh test.

• Keep the real probe flow and unchanged 1,500 ms production deadline.
• Preserve deadline propagation, healthy-sibling publication, provider separation, and discovery cleanup assertions.
• Tighten the elapsed-time assertion from a wall-clock tolerance to the exact virtual deadline.
• No production code or new dependencies. Most of the diff is indentation.

Measured benefit

Same test, forced uncached with -count=1, measured using go test -json:

Measurement	Before	After
Target test	1.50 s	<0.01 s
Selected package run	1.542 s	0.045 s

The post-change test duration rounds to zero at the JSON reporter's precision; this is not a whole-suite speedup claim.

Completed verification

• Targeted before/after test runs passed.
• Probe, refresh-deadline, and concurrent-refresh tests passed 50 repetitions under -race.
• Mutation check: adding 1 ms to the effective bootstrap budget failed with probe elapsed = 1.501s, want 1.5s. The mutation was reverted and the test passed again.
• Full non-race task test GOPROXY=off passed.
• Independent focused review found no blockers or important findings.
• git diff --check passed.

Outstanding verification

Opening ready for review at the operator's request, with incomplete aggregate checks disclosed:

• The initial root lint scan reported zero issues but exceeded its timeout.
• The retry passed root lint; engine/submodule lint completion has not yet been confirmed.
• Full repository race verification and the offline demo have not yet completed.

These are not claimed as passed. CI checks remain pending on the submitted branch.

Classification

Routine: test-only use of existing standard-library virtual time. No behavior, API, configuration, or documentation changes.