test: convert pty_session wall-clock waits to deterministic predicate witnesses

[Kyzcreig]

Defect

tests/hermes_cli/test_pty_session.py::test_reaper_loop_invokes_reap is a wall-clock flake. It starts the reaper at interval=0.01, sleeps 0.05, then asserts the reap counter >= 2:

task = asyncio.create_task(run_reaper(reg, interval=0.01))
await asyncio.sleep(0.05)
task.cancel()
...
assert calls["n"] >= 2

On a loaded CI runner that stopwatch measures the machine, not the reaper. We hit assert 1 >= 2 on a PR that touches no pty code at all (agent/credential_pool.py only) — the test reds out other people's work.

Conversion

Tests only — hermes_cli/pty_session.py is untouched; no production seam was needed.

• test_reaper_loop_invokes_reap now awaits an asyncio.Event the fake reap_idle sets on its second call, under a 10s wait_for hang guard. The witness is the iteration count reached, so load costs time, never correctness.
• The four sibling drain tests in the same file replace fixed asyncio.sleep(0.05) budgets with a wait_until(predicate) helper that polls the exact fact each test asserts (buffer contents / socket close code). Converting the whole family in one pass keeps the file from taxing unrelated PRs later.

The one remaining sleep(0.05) in the file is deliberately left alone: it precedes an assertion that something did not happen (s.attached is True, last_detached_at is None), so load only makes it more permissive.

RED proofs

Each mutation's red names the new witness, not an incidental error. Production restored and re-verified green after each.

mutation	red
run_reaper while True → for _ in range(1)	AssertionError: reaper loop only invoked reap_idle 1 time(s)
_drain self.buffer.append(chunk) → pass	3 failures, all timed out after 10.0s waiting for the drain task to buffer ...

Verification

$ pytest tests/hermes_cli/test_pty_session.py -q
16 passed, 2 warnings in 1.25s

Also verified under load on the equivalent file in our downstream tree: 20/20 green for the reaper test and 20/20 for the whole file with 64 CPU burners on 32 cores (load average 14 → 75).