test(hub): stop the Chrome profile cleanup racing its own children

`terminate()` signals the parent only. Chrome's zygote, renderer and gpu
children outlive it by a moment and go on writing into the profile, so rmtree
walked a directory that gained a file between its readdir and its rmdir and
raised "Directory not empty". The probe exited non-zero, and every test in the
file errored at setup — intermittently, roughly one run in three, for a reason
nowhere near the chat code they were testing.

TemporaryDirectory(ignore_cleanup_errors=True) in all four probes that own a
profile: a few bytes left in a throwaway directory are harmless, failing the run
is not. `proc.wait()` after `kill()` was also missing — a killed process still
has to be reaped.

layout_probe.py never cleaned up at all (mkdtemp, no removal) and never waited
for Chrome; it leaked a profile into /tmp on every run.

Ten consecutive runs of test_chat_send.py are clean.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01HCGdheDLxGReuKHga3BtST

5 files changed, 50 insertions, 5 deletions

diff --git a/packages/meshbay-hub/tests/harness/chat_scroll_probe.py b/packages/meshbay-hub/tests/harness/chat_scroll_probe.py
index 7373976..17ec554 100644
--- a/packages/meshbay-hub/tests/harness/chat_scroll_probe.py
+++ b/packages/meshbay-hub/tests/harness/chat_scroll_probe.py
@@ -193,7 +193,15 @@ class H(http.server.BaseHTTPRequestHandler):
 def main() -> int:
     with socketserver.TCPServer(("127.0.0.1", PORT), H) as srv:
         threading.Thread(target=srv.serve_forever, daemon=True).start()
-        with tempfile.TemporaryDirectory() as profile:
+        # ignore_cleanup_errors: Chrome's children (zygote, renderer, gpu)
+        # outlive terminate() on the parent by a moment and go on writing into
+        # the profile. rmtree then walks a directory that gains a file between
+        # its readdir and its rmdir and raises "Directory not empty" -- which
+        # failed the probe, which failed every test in the file, intermittently
+        # and for a reason nowhere near the chat code they were testing. A few
+        # bytes left in a throwaway profile are harmless; failing the run is not.
+        with tempfile.TemporaryDirectory(
+                ignore_cleanup_errors=True) as profile:
             # Real time, not `--virtual-time-budget`: the defect is a feedback
             # loop between layout and an event, and a virtual clock does not
             # run it.
@@ -211,6 +219,7 @@ def main() -> int:
                 proc.wait(timeout=10)
             except subprocess.TimeoutExpired:
                 proc.kill()
+                proc.wait()
     if not RECORDS:
         print(json.dumps({"error": "no measurement"}), file=sys.stderr)
         return 1
diff --git a/packages/meshbay-hub/tests/harness/chat_send_probe.py b/packages/meshbay-hub/tests/harness/chat_send_probe.py
index 5f99beb..28635b0 100644
--- a/packages/meshbay-hub/tests/harness/chat_send_probe.py
+++ b/packages/meshbay-hub/tests/harness/chat_send_probe.py
@@ -297,7 +297,15 @@ class H(http.server.BaseHTTPRequestHandler):
 def main() -> int:
     with socketserver.TCPServer(("127.0.0.1", PORT), H) as srv:
         threading.Thread(target=srv.serve_forever, daemon=True).start()
-        with tempfile.TemporaryDirectory() as profile:
+        # ignore_cleanup_errors: Chrome's children (zygote, renderer, gpu)
+        # outlive terminate() on the parent by a moment and go on writing into
+        # the profile. rmtree then walks a directory that gains a file between
+        # its readdir and its rmdir and raises "Directory not empty" -- which
+        # failed the probe, which failed every test in the file, intermittently
+        # and for a reason nowhere near the chat code they were testing. A few
+        # bytes left in a throwaway profile are harmless; failing the run is not.
+        with tempfile.TemporaryDirectory(
+                ignore_cleanup_errors=True) as profile:
             proc = subprocess.Popen(
                 ["google-chrome", "--headless=new", "--disable-gpu", "--no-sandbox",
                  f"--user-data-dir={profile}", "--window-size=1100,800",
@@ -312,6 +320,7 @@ def main() -> int:
                 proc.wait(timeout=10)
             except subprocess.TimeoutExpired:
                 proc.kill()
+                proc.wait()
     if not RECORDS:
         print(json.dumps({"error": "no measurement"}), file=sys.stderr)
         return 1
diff --git a/packages/meshbay-hub/tests/harness/group_tab_probe.py b/packages/meshbay-hub/tests/harness/group_tab_probe.py
index 5e4e452..b6d2bcc 100644
--- a/packages/meshbay-hub/tests/harness/group_tab_probe.py
+++ b/packages/meshbay-hub/tests/harness/group_tab_probe.py
@@ -156,7 +156,15 @@ class H(http.server.BaseHTTPRequestHandler):
 def main() -> int:
     with socketserver.TCPServer(("127.0.0.1", PORT), H) as srv:
         threading.Thread(target=srv.serve_forever, daemon=True).start()
-        with tempfile.TemporaryDirectory() as profile:
+        # ignore_cleanup_errors: Chrome's children (zygote, renderer, gpu)
+        # outlive terminate() on the parent by a moment and go on writing into
+        # the profile. rmtree then walks a directory that gains a file between
+        # its readdir and its rmdir and raises "Directory not empty" -- which
+        # failed the probe, which failed every test in the file, intermittently
+        # and for a reason nowhere near the chat code they were testing. A few
+        # bytes left in a throwaway profile are harmless; failing the run is not.
+        with tempfile.TemporaryDirectory(
+                ignore_cleanup_errors=True) as profile:
             proc = subprocess.Popen(
                 ["google-chrome", "--headless=new", "--disable-gpu", "--no-sandbox",
                  f"--user-data-dir={profile}", "--window-size=1100,900",
@@ -171,6 +179,7 @@ def main() -> int:
                 proc.wait(timeout=10)
             except subprocess.TimeoutExpired:
                 proc.kill()
+                proc.wait()
     if not RECORDS:
         print(json.dumps({"error": "no measurement"}), file=sys.stderr)
         return 1
diff --git a/packages/meshbay-hub/tests/harness/layout_probe.py b/packages/meshbay-hub/tests/harness/layout_probe.py
index 530b3f0..65b3083 100644
--- a/packages/meshbay-hub/tests/harness/layout_probe.py
+++ b/packages/meshbay-hub/tests/harness/layout_probe.py
@@ -19,6 +19,7 @@ single pass. Launching Chrome per width put three minutes on the test suite.
 """
 import http.server
 import json
+import shutil
 import socketserver
 import subprocess
 import sys
@@ -118,16 +119,25 @@ def main() -> int:
 
     srv = S(("127.0.0.1", PORT), H)
     threading.Thread(target=srv.serve_forever, daemon=True).start()
+    # mkdtemp left a Chrome profile in /tmp on every run, for ever, and nothing
+    # waited for Chrome to exit. Same cleanup rule as the other probes.
+    profile = tempfile.mkdtemp(prefix="chrome-layout-")
     chrome = subprocess.Popen([
         "google-chrome", "--headless=new", "--no-sandbox",
         "--window-size=1000,900",
-        "--user-data-dir=" + tempfile.mkdtemp(prefix="chrome-layout-"),
+        "--user-data-dir=" + profile,
         f"http://127.0.0.1:{PORT}/",
     ], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
     deadline = time.time() + 45
     while time.time() < deadline and not RECORDS:
         time.sleep(0.2)
     chrome.terminate()
+    try:
+        chrome.wait(timeout=10)
+    except subprocess.TimeoutExpired:
+        chrome.kill()
+        chrome.wait()
+    shutil.rmtree(profile, ignore_errors=True)
     srv.shutdown()
     if not RECORDS:
         print(json.dumps({"error": "no measurement"}))
diff --git a/packages/meshbay-hub/tests/harness/scroll_probe.py b/packages/meshbay-hub/tests/harness/scroll_probe.py
index 46d8357..ae407b8 100644
--- a/packages/meshbay-hub/tests/harness/scroll_probe.py
+++ b/packages/meshbay-hub/tests/harness/scroll_probe.py
@@ -151,7 +151,15 @@ class H(http.server.BaseHTTPRequestHandler):
 def main() -> int:
     with socketserver.TCPServer(("127.0.0.1", PORT), H) as srv:
         threading.Thread(target=srv.serve_forever, daemon=True).start()
-        with tempfile.TemporaryDirectory() as profile:
+        # ignore_cleanup_errors: Chrome's children (zygote, renderer, gpu)
+        # outlive terminate() on the parent by a moment and go on writing into
+        # the profile. rmtree then walks a directory that gains a file between
+        # its readdir and its rmdir and raises "Directory not empty" -- which
+        # failed the probe, which failed every test in the file, intermittently
+        # and for a reason nowhere near the chat code they were testing. A few
+        # bytes left in a throwaway profile are harmless; failing the run is not.
+        with tempfile.TemporaryDirectory(
+                ignore_cleanup_errors=True) as profile:
             subprocess.run(
                 ["google-chrome", "--headless", "--disable-gpu", "--no-sandbox",
                  f"--user-data-dir={profile}", "--window-size=1100,1300",