diff --git a/agent/deadline.py b/agent/deadline.py
index 2be255c744..83d7f3cf2c 100644
--- a/agent/deadline.py
+++ b/agent/deadline.py
@@ -71,7 +71,7 @@ import sys
 import threading
 import time
 from dataclasses import dataclass
-from typing import Any, Protocol, Awaitable, Callable, Optional
+from typing import Any, Awaitable, Callable, Optional, Protocol
 
 logger = logging.getLogger(__name__)
 
@@ -126,6 +126,12 @@ class SuspectableBackend(Protocol):
     instead of returning a poisoned handle to the cache. Consumers adopt
     incrementally (Phase 3b, one backend per PR), so the layer fails open:
     backends without the protocol are simply never marked.
+
+    Adopter contract: ``mark_suspect`` MUST be cheap, non-blocking, and
+    must not acquire locks the guarded operation may hold. It runs inline —
+    on the event loop in the async flavor, and on the caller's thread in
+    the sync flavor while the wedged worker is still alive. Set a flag;
+    do the expensive health-check/recycle work in ``ensure_healthy``.
     """
 
     def mark_suspect(self, reason: str) -> None: ...
@@ -426,6 +432,11 @@ async def run_bounded_async(
             cleanup.add_done_callback(_consume_abandoned)
         # Phase 3a (#85125): the abandoned task may leave the backend
         # half-wedged; flag it so the owner recycles before reuse.
+        # Deliberately INLINE on the loop (adopter contract: mark_suspect is
+        # cheap and non-blocking). Running it synchronously guarantees the
+        # mark happens-before this BoundedResult returns AND before the
+        # ensure_future'd on_abandon cleanup can start (next loop tick) — an
+        # offloaded mark would race both.
         _mark_backend_suspect(backend, label, timeout_s)
         logger.warning(
             "[deadline] %r timed out after %.1fs; task abandoned", label, timeout_s
@@ -507,8 +518,7 @@ def run_bounded_sync(
         # recycle/re-init in on_timeout never gets a stale flag on the healed
         # replacement. The sync flavor runs the mark inline — the protocol
         # contract requires mark_suspect to be cheap.
-        if backend is not None:
-            _mark_backend_suspect(backend, label, timeout_s)
+        _mark_backend_suspect(backend, label, timeout_s)
         if on_timeout is not None:
             try:
                 on_timeout()
diff --git a/tests/agent/test_deadline.py b/tests/agent/test_deadline.py
index 6817bad16f..e66ac19b2e 100644
--- a/tests/agent/test_deadline.py
+++ b/tests/agent/test_deadline.py
@@ -575,7 +575,7 @@ def test_async_timeout_marks_backend_once():
     backend = asyncio.run(drive())
     assert len(backend.reasons) == 1
     assert "phase3a" in backend.reasons[0]
-    assert "0.1" in backend.reasons[0]  # rounded timeout present
+    assert "timed out after 0.1" in backend.reasons[0]
 
 
 def test_async_completion_never_marks_backend():
@@ -650,3 +650,68 @@ def test_mark_suspect_raising_never_corrupts_the_result():
     result = asyncio.run(drive())
     assert result.timed_out
     assert result.label == "phase3a-boom"
+
+
+def test_sync_completion_never_marks_backend():
+    from agent.deadline import run_bounded_sync
+
+    backend = _RecordingBackend()
+    result = run_bounded_sync(
+        lambda: "ok", 5.0, label="phase3a-sync-ok", backend=backend
+    )
+    assert not result.timed_out and result.value == "ok"
+    assert backend.reasons == []
+
+
+def test_sync_mark_happens_before_on_timeout():
+    """The review-round ordering contract: mark BEFORE owner cleanup, so a
+    recycle in on_timeout never sees an unmarked backend (and a healed
+    replacement never inherits a stale flag)."""
+    from agent.deadline import run_bounded_sync
+
+    backend = _RecordingBackend()
+    seen_at_cleanup: list[int] = []
+
+    def on_timeout():
+        seen_at_cleanup.append(len(backend.reasons))
+
+    result = run_bounded_sync(
+        lambda: time.sleep(10),
+        0.05,
+        label="phase3a-order-sync",
+        on_timeout=on_timeout,
+        backend=backend,
+    )
+    assert result.timed_out
+    assert seen_at_cleanup == [1]  # mark already applied when cleanup ran
+
+
+def test_async_mark_happens_before_on_abandon_cleanup():
+    """Pins the scheduling invariant the inline mark relies on: on_abandon
+    is ensure_future'd (can't start until the next loop tick), so the
+    synchronous mark always lands first. An offloaded (to_thread) mark
+    would break this — this test is the guard against that 'fix'."""
+    from agent.deadline import run_bounded_async
+
+    backend = _RecordingBackend()
+    seen_at_cleanup: list[int] = []
+    cleaned = asyncio.Event()
+
+    async def _cleanup():
+        seen_at_cleanup.append(len(backend.reasons))
+        cleaned.set()
+
+    async def never():
+        await asyncio.Event().wait()
+
+    async def drive():
+        result = await run_bounded_async(
+            never(), 0.05, label="phase3a-order-async",
+            on_abandon=_cleanup, backend=backend,
+        )
+        await asyncio.wait_for(cleaned.wait(), timeout=5.0)
+        return result
+
+    result = asyncio.run(drive())
+    assert result.timed_out
+    assert seen_at_cleanup == [1]  # mark already applied when cleanup started