From 9baf44f0db14cb427de3a911665d061fb1bfff00 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:03:10 +0900
Subject: [PATCH 01/52] test: require Flask Gunicorn gthread profile

---
 tests/test_python_flask_acceptance.py | 11 +++++++----
 1 file changed, 7 insertions(+), 4 deletions(-)

diff --git a/tests/test_python_flask_acceptance.py b/tests/test_python_flask_acceptance.py
index d773850..62b50e8 100644
--- a/tests/test_python_flask_acceptance.py
+++ b/tests/test_python_flask_acceptance.py
@@ -13,14 +13,17 @@ def test_approved_wsgi_profile_is_pinned(self):
         self.assertEqual((APP / ".python-version").read_text().strip(), "3.14.7")
         requirements = set((APP / "requirements.in").read_text().splitlines())
         self.assertIn("flask==3.1.3", requirements)
-        self.assertIn("waitress==3.0.2", requirements)
+        self.assertIn("gunicorn==26.0.0", requirements)
+        self.assertNotIn("waitress==3.0.2", requirements)
         self.assertIn("psycopg==3.3.5", requirements)
         self.assertIn("psycopg-binary==3.3.5", requirements)
         self.assertIn("psycopg-pool==3.3.1", requirements)
         server = (APP / "benchmark_api" / "server.py").read_text()
-        self.assertIn("threads=1", server)
-        self.assertIn("create_server", server)
-        self.assertNotIn("gunicorn", server.lower())
+        self.assertIn('workers = 1', server)
+        self.assertIn('threads = 1', server)
+        self.assertIn('worker_class = "gthread"', server)
+        self.assertIn('keepalive = 5', server)
+        self.assertNotIn("waitress", server.lower())
 
     def test_pool_and_container_limits_are_not_weakened(self):
         database = (APP / "benchmark_api" / "database.py").read_text()

From 9cf93dc3c9b12804a30f66c5411e483565915fe3 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:03:15 +0900
Subject: [PATCH 02/52] test: prove Flask Gunicorn acceptance red

---
 .github/workflows/flask-gunicorn-tdd.yml | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
 create mode 100644 .github/workflows/flask-gunicorn-tdd.yml

diff --git a/.github/workflows/flask-gunicorn-tdd.yml b/.github/workflows/flask-gunicorn-tdd.yml
new file mode 100644
index 0000000..3ba7e2f
--- /dev/null
+++ b/.github/workflows/flask-gunicorn-tdd.yml
@@ -0,0 +1,15 @@
+name: Temporary Flask Gunicorn TDD red
+on:
+  push:
+    branches: [fix/flask-gunicorn-70]
+permissions:
+  contents: read
+jobs:
+  red:
+    runs-on: ubuntu-24.04
+    steps:
+      - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09
+      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
+        with:
+          python-version-file: apps/python-flask/.python-version
+      - run: python -m unittest tests.test_python_flask_acceptance

From b10c85c98fe1ff847dc60cde3dcf53df47b3b65e Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:04:16 +0900
Subject: [PATCH 03/52] test: define Gunicorn process validation

---
 tests/test_benchmark_environment.py | 28 ++++++++++++++++++++++++++++
 1 file changed, 28 insertions(+)

diff --git a/tests/test_benchmark_environment.py b/tests/test_benchmark_environment.py
index 3041e21..1129020 100644
--- a/tests/test_benchmark_environment.py
+++ b/tests/test_benchmark_environment.py
@@ -163,6 +163,34 @@ def execute(arguments, *, timeout, tick=None, cwd=environment.ROOT):
             self.assertEqual(result["memory_samples"], 1)
 
 
+class ProcessValidationTests(unittest.TestCase):
+    def test_flask_gunicorn_allows_exactly_one_arbiter_and_one_worker(self):
+        value = state()
+        value["Config"]["Labels"]["com.docker.compose.service"] = "python-flask"
+        value["Path"] = "python"
+        value["Args"] = ["-m", "benchmark_api.server"]
+        output = (
+            "PID COMMAND\n"
+            "10 python -m benchmark_api.server\n"
+            "11 python -m benchmark_api.server\n"
+        )
+        environment.validate_processes(value, output, allow_health_probe=False)
+
+    def test_flask_gunicorn_rejects_extra_worker(self):
+        value = state()
+        value["Config"]["Labels"]["com.docker.compose.service"] = "python-flask"
+        value["Path"] = "python"
+        value["Args"] = ["-m", "benchmark_api.server"]
+        output = (
+            "PID COMMAND\n"
+            "10 python -m benchmark_api.server\n"
+            "11 python -m benchmark_api.server\n"
+            "12 python -m benchmark_api.server\n"
+        )
+        with self.assertRaises(BenchmarkFailure):
+            environment.validate_processes(value, output, allow_health_probe=False)
+
+
 class InstallerTests(unittest.TestCase):
     def test_cached_checksum_is_checked_before_executing_binary(self):
         with tempfile.TemporaryDirectory() as directory:

From 0143d380ec1a5e1e7621eabcbca237465556f613 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:04:23 +0900
Subject: [PATCH 04/52] test: cover Gunicorn process validation red

---
 .github/workflows/flask-gunicorn-tdd.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/flask-gunicorn-tdd.yml b/.github/workflows/flask-gunicorn-tdd.yml
index 3ba7e2f..79c4e21 100644
--- a/.github/workflows/flask-gunicorn-tdd.yml
+++ b/.github/workflows/flask-gunicorn-tdd.yml
@@ -12,4 +12,4 @@ jobs:
       - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
         with:
           python-version-file: apps/python-flask/.python-version
-      - run: python -m unittest tests.test_python_flask_acceptance
+      - run: python -m unittest tests.test_python_flask_acceptance tests.test_benchmark_environment

From 7d0e54b0758fe1a30d708d81449db1465bf66067 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:04:45 +0900
Subject: [PATCH 05/52] test: assert Gunicorn runtime profile at container
 boundary

---
 tests/test_python_flask_acceptance.py | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/tests/test_python_flask_acceptance.py b/tests/test_python_flask_acceptance.py
index 62b50e8..2695aa3 100644
--- a/tests/test_python_flask_acceptance.py
+++ b/tests/test_python_flask_acceptance.py
@@ -18,12 +18,12 @@ def test_approved_wsgi_profile_is_pinned(self):
         self.assertIn("psycopg==3.3.5", requirements)
         self.assertIn("psycopg-binary==3.3.5", requirements)
         self.assertIn("psycopg-pool==3.3.1", requirements)
-        server = (APP / "benchmark_api" / "server.py").read_text()
-        self.assertIn('workers = 1', server)
-        self.assertIn('threads = 1', server)
-        self.assertIn('worker_class = "gthread"', server)
-        self.assertIn('keepalive = 5', server)
-        self.assertNotIn("waitress", server.lower())
+        dockerfile = (APP / "Dockerfile").read_text()
+        self.assertIn('"--workers", "1"', dockerfile)
+        self.assertIn('"--threads", "1"', dockerfile)
+        self.assertIn('"--worker-class", "gthread"', dockerfile)
+        self.assertIn('"--keep-alive", "5"', dockerfile)
+        self.assertNotIn("waitress", dockerfile.lower())
 
     def test_pool_and_container_limits_are_not_weakened(self):
         database = (APP / "benchmark_api" / "database.py").read_text()
@@ -31,7 +31,7 @@ def test_pool_and_container_limits_are_not_weakened(self):
         dockerfile = (APP / "Dockerfile").read_text()
         self.assertEqual(dockerfile.count("python:3.14.7-slim-bookworm@sha256:"), 2)
         self.assertIn("USER 10001:10001", dockerfile)
-        self.assertIn('ENTRYPOINT ["python", "-m", "benchmark_api.server"]', dockerfile)
+        self.assertIn('ENTRYPOINT ["python", "-m", "gunicorn.app.wsgiapp"]', dockerfile)
 
     def test_registration_preserves_the_frozen_legacy_cohort(self):
         registry = json.loads((ROOT / "benchmark" / "implementations.json").read_text())

From 23438846b1f49f3b2d1c53a269eb35ccb24fc63c Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:04:54 +0900
Subject: [PATCH 06/52] feat: replace Waitress with Gunicorn

---
 apps/python-flask/requirements.in | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/apps/python-flask/requirements.in b/apps/python-flask/requirements.in
index 691945c..a3e98fe 100644
--- a/apps/python-flask/requirements.in
+++ b/apps/python-flask/requirements.in
@@ -1,5 +1,5 @@
 flask==3.1.3
-waitress==3.0.2
+gunicorn==26.0.0
 psycopg==3.3.5
 psycopg-binary==3.3.5
 psycopg-pool==3.3.1

From 908797def38c1dc50519ae4ee4bb58d994b3071c Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:04:57 +0900
Subject: [PATCH 07/52] feat: add Gunicorn Flask application factory

---
 apps/python-flask/benchmark_api/server.py | 44 ++++-------------------
 1 file changed, 6 insertions(+), 38 deletions(-)

diff --git a/apps/python-flask/benchmark_api/server.py b/apps/python-flask/benchmark_api/server.py
index 9a264a1..2b018f6 100644
--- a/apps/python-flask/benchmark_api/server.py
+++ b/apps/python-flask/benchmark_api/server.py
@@ -1,45 +1,13 @@
-"""Single-process Waitress entry point with explicit graceful resource cleanup."""
+"""Gunicorn application factory with worker-local database resources."""
 
-import signal
-
-from waitress.server import create_server
+import atexit
 
 from benchmark_api.app import create_app
 from benchmark_api.database import close_pool, open_pool
 
 
-def serve() -> None:
+def create_server_app():
+    """Create one worker-local Flask app and close its pool when the worker exits."""
     pool = open_pool()
-    server = None
-    previous = {}
-    try:
-        app = create_app(pool)
-        server = create_server(
-            app,
-            host="0.0.0.0",
-            port=8080,
-            threads=1,
-            channel_timeout=30,
-            cleanup_interval=10,
-            asyncore_use_poll=True,
-        )
-
-        def stop(_signum, _frame):
-            server.close()
-
-        for sig in (signal.SIGINT, signal.SIGTERM):
-            previous[sig] = signal.signal(sig, stop)
-        server.run()
-    finally:
-        for sig, handler in previous.items():
-            signal.signal(sig, handler)
-        if server is not None:
-            server.close()
-            dispatcher = getattr(server, "task_dispatcher", None)
-            if dispatcher is not None:
-                dispatcher.shutdown(cancel_pending=False, timeout=5)
-        close_pool(pool)
-
-
-if __name__ == "__main__":
-    serve()
+    atexit.register(close_pool, pool)
+    return create_app(pool)

From a9dbe4405c5dc78be312eb1510775ead8110b954 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:04:59 +0900
Subject: [PATCH 08/52] feat: run Flask with one Gunicorn gthread worker

---
 apps/python-flask/Dockerfile | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/apps/python-flask/Dockerfile b/apps/python-flask/Dockerfile
index a5c61dd..443c516 100644
--- a/apps/python-flask/Dockerfile
+++ b/apps/python-flask/Dockerfile
@@ -16,4 +16,5 @@ COPY --from=dependencies /opt/venv /opt/venv
 COPY benchmark_api ./benchmark_api
 USER 10001:10001
 EXPOSE 8080
-ENTRYPOINT ["python", "-m", "benchmark_api.server"]
+ENTRYPOINT ["python", "-m", "gunicorn.app.wsgiapp"]
+CMD ["--bind", "0.0.0.0:8080", "--workers", "1", "--worker-class", "gthread", "--threads", "1", "--keep-alive", "5", "--timeout", "30", "--graceful-timeout", "5", "--worker-tmp-dir", "/dev/shm", "benchmark_api.server:create_server_app()"]

From 9bb34ead5b3ff4a83e30b56ddec496ba75e1598e Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:05:06 +0900
Subject: [PATCH 09/52] chore: record Gunicorn in Flask version metadata

---
 benchmark/implementations.json | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/benchmark/implementations.json b/benchmark/implementations.json
index 6b66854..89f6be5 100644
--- a/benchmark/implementations.json
+++ b/benchmark/implementations.json
@@ -9,7 +9,7 @@
     {"id":"node-fastify","language":"Node.js","framework":"Fastify","display_name":"Node.js / Fastify","source_path":"apps/node-fastify","version_fields":["node","fastify","pg"],"acceptance_test":"tests/test_node_fastify_service.py","failure_test":"tests/test_node_fastify_acceptance.py"},
     {"id":"node-express","language":"Node.js","framework":"Express","display_name":"Node.js / Express","source_path":"apps/node-express","version_fields":["node","express","pg"],"acceptance_test":"tests/test_node_express_service.py","failure_test":"tests/test_node_express_acceptance.py"},
     {"id":"python-fastapi","language":"Python","framework":"FastAPI","display_name":"Python / FastAPI","source_path":"apps/python-fastapi","version_fields":["python","fastapi","uvicorn","asyncpg"],"acceptance_test":"tests/test_python_fastapi_service.py","failure_test":"tests/test_python_fastapi_acceptance.py"},
-    {"id":"python-flask","language":"Python","framework":"Flask","display_name":"Python / Flask","source_path":"apps/python-flask","version_fields":["python","flask","waitress","psycopg","psycopg-binary","psycopg-pool"],"acceptance_test":"tests/test_python_flask_service.py","failure_test":"tests/test_python_flask_acceptance.py"},
+    {"id":"python-flask","language":"Python","framework":"Flask","display_name":"Python / Flask","source_path":"apps/python-flask","version_fields":["python","flask","gunicorn","psycopg","psycopg-binary","psycopg-pool"],"acceptance_test":"tests/test_python_flask_service.py","failure_test":"tests/test_python_flask_acceptance.py"},
     {"id":"java-spring-boot","language":"Java","framework":"Spring Boot","display_name":"Java / Spring Boot","source_path":"apps/java-spring-boot","version_fields":["java","spring-boot","tomcat","postgresql","hikaricp","jackson","gradle"],"acceptance_test":"tests/test_java_spring_boot_service.py","failure_test":"tests/test_java_spring_boot_acceptance.py"}
   ],
   "cohorts": {

From eb8d434ef4d90eabd796bc43b684a7fb71893ecd Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:05:09 +0900
Subject: [PATCH 10/52] fix: validate Flask Gunicorn arbiter and worker

---
 benchmark/environment.py | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)

diff --git a/benchmark/environment.py b/benchmark/environment.py
index f18e66f..e28f1b6 100644
--- a/benchmark/environment.py
+++ b/benchmark/environment.py
@@ -128,12 +128,14 @@ def normalized(arguments):
         pieces = row.strip().split(None, 1)
         require(len(pieces) == 2 and pieces[0].isdigit(), "invalid process list row")
         commands.append(normalized(shlex.split(pieces[1])))
+    service = state["Config"]["Labels"]["com.docker.compose.service"]
+    expected_servers = 2 if service == "python-flask" else 1
     require(
-        commands.count(server) == 1 and all(command in allowed for command in commands),
+        commands.count(server) == expected_servers and all(command in allowed for command in commands),
         (
-            "expected one server process and only independent health probes"
+            f"expected {expected_servers} server process(es) and only independent health probes"
             if allow_health_probe
-            else "expected exactly one server process"
+            else f"expected exactly {expected_servers} server process(es)"
         ),
     )
 

From 019b5b29b394f15314e46d5ea2bd4bf6fae4f7d1 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:05:12 +0900
Subject: [PATCH 11/52] test: validate Gunicorn runtime process model

---
 tests/test_python_flask_service.py | 18 ++++++++++++++----
 1 file changed, 14 insertions(+), 4 deletions(-)

diff --git a/tests/test_python_flask_service.py b/tests/test_python_flask_service.py
index 47ae42e..0110b83 100644
--- a/tests/test_python_flask_service.py
+++ b/tests/test_python_flask_service.py
@@ -121,8 +121,15 @@ def running_service() -> str:
     require(state["HostConfig"]["Memory"] == 536_870_912, "memory limit differs")
     require(state["Config"]["User"] == "10001:10001", "runtime user differs")
     require(
-        state["Path"] == "python" and state["Args"] == ["-m", "benchmark_api.server"],
-        "unexpected server wrapper",
+        state["Path"] == "python"
+        and state["Args"][:2] == ["-m", "gunicorn.app.wsgiapp"]
+        and "--workers" in state["Args"]
+        and state["Args"][state["Args"].index("--workers") + 1] == "1"
+        and "--worker-class" in state["Args"]
+        and state["Args"][state["Args"].index("--worker-class") + 1] == "gthread"
+        and "--threads" in state["Args"]
+        and state["Args"][state["Args"].index("--threads") + 1] == "1",
+        "unexpected Gunicorn runtime profile",
     )
     require(state["HostConfig"]["CapDrop"] == ["ALL"], "capabilities not dropped")
     require(
@@ -133,8 +140,11 @@ def running_service() -> str:
         ["docker", "top", container, "-eo", "pid,args"]
     ).stdout.splitlines()[1:]
     commands = [line.split(maxsplit=1)[1] for line in processes if line.strip()]
-    servers = [command for command in commands if "benchmark_api.server" in command]
-    require(len(servers) == 1, f"expected one Waitress server process, got {servers!r}")
+    servers = [command for command in commands if "gunicorn" in command]
+    require(
+        len(servers) == 2,
+        f"expected one Gunicorn arbiter and one request worker, got {servers!r}",
+    )
     return container
 
 

From a7916f5ffeae6e91ee60a92907f939a1afab458a Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:05:20 +0900
Subject: [PATCH 12/52] test: verify Gunicorn profile and generate lock

---
 .github/workflows/flask-gunicorn-tdd.yml | 16 +++++++++++++---
 1 file changed, 13 insertions(+), 3 deletions(-)

diff --git a/.github/workflows/flask-gunicorn-tdd.yml b/.github/workflows/flask-gunicorn-tdd.yml
index 79c4e21..36ecb0a 100644
--- a/.github/workflows/flask-gunicorn-tdd.yml
+++ b/.github/workflows/flask-gunicorn-tdd.yml
@@ -1,15 +1,25 @@
-name: Temporary Flask Gunicorn TDD red
+name: Temporary Flask Gunicorn TDD
 on:
   push:
     branches: [fix/flask-gunicorn-70]
 permissions:
   contents: read
 jobs:
-  red:
+  verify:
     runs-on: ubuntu-24.04
     steps:
       - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09
       - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
         with:
           python-version-file: apps/python-flask/.python-version
-      - run: python -m unittest tests.test_python_flask_acceptance tests.test_benchmark_environment
+      - name: Unit acceptance
+        run: python -m unittest tests.test_python_flask_acceptance tests.test_benchmark_environment
+      - name: Generate runtime lock
+        run: |
+          python -m pip install pip-tools==7.5.2
+          python -m piptools compile --generate-hashes --output-file=apps/python-flask/requirements.lock apps/python-flask/requirements.in
+      - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
+        with:
+          name: flask-gunicorn-lock
+          path: apps/python-flask/requirements.lock
+          if-no-files-found: error

From 6c01f3a32de2a838cb3aa61b946a6b68bbab3413 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:05:49 +0900
Subject: [PATCH 13/52] fix: preserve generic process validation fixtures

---
 benchmark/environment.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/benchmark/environment.py b/benchmark/environment.py
index e28f1b6..160e269 100644
--- a/benchmark/environment.py
+++ b/benchmark/environment.py
@@ -128,7 +128,7 @@ def normalized(arguments):
         pieces = row.strip().split(None, 1)
         require(len(pieces) == 2 and pieces[0].isdigit(), "invalid process list row")
         commands.append(normalized(shlex.split(pieces[1])))
-    service = state["Config"]["Labels"]["com.docker.compose.service"]
+    service = state.get("Config", {}).get("Labels", {}).get("com.docker.compose.service")
     expected_servers = 2 if service == "python-flask" else 1
     require(
         commands.count(server) == expected_servers and all(command in allowed for command in commands),

From b2f7f9d61962ee12bd80a59424e79bb65b0f5006 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:06:20 +0900
Subject: [PATCH 14/52] test: pin pip for runtime lock generation

---
 .github/workflows/flask-gunicorn-tdd.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/flask-gunicorn-tdd.yml b/.github/workflows/flask-gunicorn-tdd.yml
index 36ecb0a..706e388 100644
--- a/.github/workflows/flask-gunicorn-tdd.yml
+++ b/.github/workflows/flask-gunicorn-tdd.yml
@@ -16,7 +16,7 @@ jobs:
         run: python -m unittest tests.test_python_flask_acceptance tests.test_benchmark_environment
       - name: Generate runtime lock
         run: |
-          python -m pip install pip-tools==7.5.2
+          python -m pip install pip==25.3 pip-tools==7.5.2
           python -m piptools compile --generate-hashes --output-file=apps/python-flask/requirements.lock apps/python-flask/requirements.in
       - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
         with:

From 4b220540c33c3514aa3c855e9db471a3b146cd7b Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:07:11 +0900
Subject: [PATCH 15/52] test: commit generated Flask runtime lock

---
 .github/workflows/flask-gunicorn-tdd.yml | 16 ++++++++++------
 1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/.github/workflows/flask-gunicorn-tdd.yml b/.github/workflows/flask-gunicorn-tdd.yml
index 706e388..36a8d6e 100644
--- a/.github/workflows/flask-gunicorn-tdd.yml
+++ b/.github/workflows/flask-gunicorn-tdd.yml
@@ -3,7 +3,7 @@ on:
   push:
     branches: [fix/flask-gunicorn-70]
 permissions:
-  contents: read
+  contents: write
 jobs:
   verify:
     runs-on: ubuntu-24.04
@@ -18,8 +18,12 @@ jobs:
         run: |
           python -m pip install pip==25.3 pip-tools==7.5.2
           python -m piptools compile --generate-hashes --output-file=apps/python-flask/requirements.lock apps/python-flask/requirements.in
-      - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
-        with:
-          name: flask-gunicorn-lock
-          path: apps/python-flask/requirements.lock
-          if-no-files-found: error
+      - name: Commit generated runtime lock
+        run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+          git add apps/python-flask/requirements.lock
+          if ! git diff --cached --quiet; then
+            git commit -m "chore: refresh Flask runtime lock"
+            git push origin HEAD:fix/flask-gunicorn-70
+          fi

From 6a9d86498b32d475c785588ad6e9837533947923 Mon Sep 17 00:00:00 2001
From: "github-actions[bot]"
 <41898282+github-actions[bot]@users.noreply.github.com>
Date: Sun, 27 Sep 2026 02:07:25 +0000
Subject: [PATCH 16/52] chore: refresh Flask runtime lock

---
 apps/python-flask/requirements.lock | 14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff --git a/apps/python-flask/requirements.lock b/apps/python-flask/requirements.lock
index b5be15f..46f0d5f 100644
--- a/apps/python-flask/requirements.lock
+++ b/apps/python-flask/requirements.lock
@@ -2,7 +2,7 @@
 # This file is autogenerated by pip-compile with Python 3.14
 # by the following command:
 #
-#    pip-compile --generate-hashes --no-index --output-file=apps/python-flask/requirements.lock --strip-extras apps/python-flask/requirements.in
+#    pip-compile --generate-hashes --no-index --output-file=apps/python-flask/requirements.lock apps/python-flask/requirements.in
 #
 blinker==1.9.0 \
     --hash=sha256:b4ce2265a7abece45e7cc896e98dbebe6cead56bcf805a3d23136d145f5445bf \
@@ -16,6 +16,10 @@ flask==3.1.3 \
     --hash=sha256:0ef0e52b8a9cd932855379197dd8f94047b359ca0a78695144304cb45f87c9eb \
     --hash=sha256:f4bcbefc124291925f1a26446da31a5178f9483862233b23c0c96a20701f670c
     # via -r apps/python-flask/requirements.in
+gunicorn==26.0.0 \
+    --hash=sha256:40233d26a5f0d1872916188c276e21641155111c2853f0c2cd55260aec0d24fc \
+    --hash=sha256:ca9346f85e3a4aeeb64d491045c16b9a35647abd37ea15efe53080eb8b090baf
+    # via -r apps/python-flask/requirements.in
 itsdangerous==2.2.0 \
     --hash=sha256:c6242fc49e35958c8b15141343aa660db5fc54d4f13a1db01a3f5891b98700ef \
     --hash=sha256:e0050c0b7da1eea53ffaf149c0cfbb5c6e2e2b69c4bef22c81fa6eb73e5f6173
@@ -118,6 +122,10 @@ markupsafe==3.0.3 \
     #   flask
     #   jinja2
     #   werkzeug
+packaging==26.3 \
+    --hash=sha256:94edc256424af38762eb31306eed28beb9f0efc50a8837492c9d6fd6004aed79 \
+    --hash=sha256:d7193f7c8e4e93f444fde0262bf90af30e16fa0ad0ad44cb553c87339b23cd1c
+    # via gunicorn
 psycopg==3.3.5 \
     --hash=sha256:ce5aa5cdb4f9379f00f487590e5890bfa7df9a164648c969ffa628505e21af4e \
     --hash=sha256:d0a3d9ccf5788af054cbd745278cb02401b5c312aeaafbf2c6144460aec47da4
@@ -187,10 +195,6 @@ typing-extensions==4.16.0 \
     --hash=sha256:481caa481374e813c1b176ada14e97f1f67a4539ce9cfeb3f350d78d6370c2e8 \
     --hash=sha256:dc983d19a509c94dba722ee6abd33940f7c05a89e243c47e907eb4db6f1a43e5
     # via psycopg-pool
-waitress==3.0.2 \
-    --hash=sha256:682aaaf2af0c44ada4abfb70ded36393f0e307f4ab9456a215ce0020baefc31f \
-    --hash=sha256:c56d67fd6e87c2ee598b76abdd4e96cfad1f24cacdea5078d382b1f9d7b5ed2e
-    # via -r apps/python-flask/requirements.in
 werkzeug==3.1.8 \
     --hash=sha256:63a77fb8892bf28ebc3178683445222aa500e48ebad5ec77b0ad80f8726b1f50 \
     --hash=sha256:9bad61a4268dac112f1c5cd4630a56ede601b6ed420300677a869083d70a4c44

From cb2407e5909dfef7d57a4aff6012c9aafbc40333 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:07:44 +0900
Subject: [PATCH 17/52] test: cover Gunicorn worker-local Flask factory

---
 apps/python-flask/tests/test_server.py | 33 +++++++++++---------------
 1 file changed, 14 insertions(+), 19 deletions(-)

diff --git a/apps/python-flask/tests/test_server.py b/apps/python-flask/tests/test_server.py
index ad148a4..a1c3a79 100644
--- a/apps/python-flask/tests/test_server.py
+++ b/apps/python-flask/tests/test_server.py
@@ -1,30 +1,25 @@
-"""Waitress lifecycle tests for the approved one-process, one-thread profile."""
+"""Gunicorn application-factory lifecycle tests."""
 
 from unittest.mock import MagicMock
 
 from benchmark_api import server
 
 
-def test_server_uses_one_waitress_thread_and_closes_resources(monkeypatch):
+def test_factory_opens_worker_local_pool_and_registers_cleanup(monkeypatch):
     pool = object()
-    waitress = MagicMock()
-    waitress.run.return_value = None
-    waitress.task_dispatcher = MagicMock()
-    create_server = MagicMock(return_value=waitress)
+    app = object()
+    open_pool = MagicMock(return_value=pool)
     close_pool = MagicMock()
-    monkeypatch.setattr(server, "open_pool", MagicMock(return_value=pool))
+    create_app = MagicMock(return_value=app)
+    register = MagicMock()
+
+    monkeypatch.setattr(server, "open_pool", open_pool)
     monkeypatch.setattr(server, "close_pool", close_pool)
-    monkeypatch.setattr(server, "create_server", create_server)
-    monkeypatch.setattr(server, "create_app", MagicMock(return_value=object()))
-    monkeypatch.setattr(server.signal, "signal", MagicMock(return_value=server.signal.SIG_DFL))
+    monkeypatch.setattr(server, "create_app", create_app)
+    monkeypatch.setattr(server.atexit, "register", register)
 
-    server.serve()
+    assert server.create_server_app() is app
 
-    kwargs = create_server.call_args.kwargs
-    assert kwargs["host"] == "0.0.0.0"
-    assert kwargs["port"] == 8080
-    assert kwargs["threads"] == 1
-    waitress.run.assert_called_once_with()
-    waitress.close.assert_called()
-    waitress.task_dispatcher.shutdown.assert_called_once_with(cancel_pending=False, timeout=5)
-    close_pool.assert_called_once_with(pool)
+    open_pool.assert_called_once_with()
+    create_app.assert_called_once_with(pool)
+    register.assert_called_once_with(close_pool, pool)

From c906c75e18455c38d0eab86073c7d48f7a888fb6 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:07:47 +0900
Subject: [PATCH 18/52] chore: regenerate registry projection

---
 site/registry.mjs | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/site/registry.mjs b/site/registry.mjs
index 089a8cd..027d943 100644
--- a/site/registry.mjs
+++ b/site/registry.mjs
@@ -145,7 +145,7 @@ export const REGISTRY = {
       "version_fields": [
         "python",
         "flask",
-        "waitress",
+        "gunicorn",
         "psycopg",
         "psycopg-binary",
         "psycopg-pool"

From 9fcb674d9190a16979fcbf759a2ffac955b60bee Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:07:49 +0900
Subject: [PATCH 19/52] docs: describe Flask Gunicorn runtime

---
 docs/FLASK.md | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/docs/FLASK.md b/docs/FLASK.md
index 26b5db6..86435f6 100644
--- a/docs/FLASK.md
+++ b/docs/FLASK.md
@@ -6,13 +6,13 @@
 
 - CPython 3.14.7, matching the FastAPI baseline.
 - Flask 3.1.3.
-- Waitress 3.0.2 as the production WSGI server.
-- One server OS process and exactly one Waitress worker thread.
+- Gunicorn 26.0.0 as the production WSGI server, using one `gthread` request worker with one thread.
+- One Gunicorn arbiter plus exactly one request-processing worker and one request thread.
 - psycopg 3.3.5 with the binary implementation package pinned to the same version.
 - psycopg-pool 3.3.1 with `min_size=1`, `max_size=10`.
 - The same PostgreSQL fixture, parameterized lookup SQL, 1 CPU / 512 MiB container budget, loopback-only host publication, dropped capabilities, no-new-privileges policy, and non-root runtime used by the other implementations.
 
-Flask's development server, Gunicorn master/worker processes, gevent/event-loop monkey patching, extra middleware, response caches, and precomputed CPU results are not used.
+Flask's development server, gevent/event-loop monkey patching, extra middleware, response caches, and precomputed CPU results are not used.
 
 ## HTTP behavior
 
@@ -27,15 +27,15 @@ The implementation provides the unchanged shared contract:
 
 ## Startup and shutdown
 
-PostgreSQL pool creation and a real `SELECT 1` readiness query complete before Waitress begins accepting HTTP traffic. SIGINT and SIGTERM close the listener, Waitress drains/shuts down its single task-dispatcher thread with a bounded timeout, and the PostgreSQL pool is then closed.
+Each Gunicorn worker creates its own PostgreSQL pool and proves readiness with a real `SELECT 1` before its Flask application is returned. Gunicorn handles SIGTERM and bounded graceful worker shutdown; worker-local pool cleanup is registered with Python `atexit`.
 
 The production image starts directly with:
 
 ```text
-python -m benchmark_api.server
+python -m gunicorn.app.wsgiapp --workers 1 --worker-class gthread --threads 1 … 'benchmark_api.server:create_server_app()'
 ```
 
-There is no supervisor or second worker process.
+Gunicorn uses one arbiter process and one request-processing worker; it does not add request workers beyond the benchmark profile's single worker.
 
 ## Verification
 

From f5838979862c96216ddc13069fc4eba221417535 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:07:57 +0900
Subject: [PATCH 20/52] test: run Flask Gunicorn real-container acceptance

---
 .github/workflows/flask-gunicorn-tdd.yml | 26 ++++++++----------------
 1 file changed, 8 insertions(+), 18 deletions(-)

diff --git a/.github/workflows/flask-gunicorn-tdd.yml b/.github/workflows/flask-gunicorn-tdd.yml
index 36a8d6e..e8723c2 100644
--- a/.github/workflows/flask-gunicorn-tdd.yml
+++ b/.github/workflows/flask-gunicorn-tdd.yml
@@ -1,29 +1,19 @@
-name: Temporary Flask Gunicorn TDD
+name: Temporary Flask Gunicorn acceptance
 on:
   push:
     branches: [fix/flask-gunicorn-70]
 permissions:
-  contents: write
+  contents: read
 jobs:
-  verify:
+  acceptance:
     runs-on: ubuntu-24.04
+    timeout-minutes: 20
     steps:
       - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09
       - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
         with:
           python-version-file: apps/python-flask/.python-version
-      - name: Unit acceptance
-        run: python -m unittest tests.test_python_flask_acceptance tests.test_benchmark_environment
-      - name: Generate runtime lock
-        run: |
-          python -m pip install pip==25.3 pip-tools==7.5.2
-          python -m piptools compile --generate-hashes --output-file=apps/python-flask/requirements.lock apps/python-flask/requirements.in
-      - name: Commit generated runtime lock
-        run: |
-          git config user.name "github-actions[bot]"
-          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
-          git add apps/python-flask/requirements.lock
-          if ! git diff --cached --quiet; then
-            git commit -m "chore: refresh Flask runtime lock"
-            git push origin HEAD:fix/flask-gunicorn-70
-          fi
+      - name: Static and shared regression tests
+        run: python -m unittest tests.test_python_flask_acceptance tests.test_benchmark_environment tests.test_benchmark_registry
+      - name: Real Flask container acceptance
+        run: python tests/test_python_flask_service.py

From 908b457ec386941c0b3d433727798cef92cfbc85 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:08:22 +0900
Subject: [PATCH 21/52] test: use repository unittest discovery path

---
 .github/workflows/flask-gunicorn-tdd.yml | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/.github/workflows/flask-gunicorn-tdd.yml b/.github/workflows/flask-gunicorn-tdd.yml
index e8723c2..22ad000 100644
--- a/.github/workflows/flask-gunicorn-tdd.yml
+++ b/.github/workflows/flask-gunicorn-tdd.yml
@@ -14,6 +14,8 @@ jobs:
         with:
           python-version-file: apps/python-flask/.python-version
       - name: Static and shared regression tests
-        run: python -m unittest tests.test_python_flask_acceptance tests.test_benchmark_environment tests.test_benchmark_registry
+        run: |
+          python -m unittest tests.test_python_flask_acceptance tests.test_benchmark_environment
+          python -m unittest discover -s tests -p 'test_benchmark_registry.py'
       - name: Real Flask container acceptance
         run: python tests/test_python_flask_service.py

From 97c56241fd1c31842e3cee023fbc0968820558d2 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:09:13 +0900
Subject: [PATCH 22/52] fix: preserve historical Flask version compatibility

---
 benchmark/implementations.json | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/benchmark/implementations.json b/benchmark/implementations.json
index 89f6be5..bab355b 100644
--- a/benchmark/implementations.json
+++ b/benchmark/implementations.json
@@ -9,7 +9,7 @@
     {"id":"node-fastify","language":"Node.js","framework":"Fastify","display_name":"Node.js / Fastify","source_path":"apps/node-fastify","version_fields":["node","fastify","pg"],"acceptance_test":"tests/test_node_fastify_service.py","failure_test":"tests/test_node_fastify_acceptance.py"},
     {"id":"node-express","language":"Node.js","framework":"Express","display_name":"Node.js / Express","source_path":"apps/node-express","version_fields":["node","express","pg"],"acceptance_test":"tests/test_node_express_service.py","failure_test":"tests/test_node_express_acceptance.py"},
     {"id":"python-fastapi","language":"Python","framework":"FastAPI","display_name":"Python / FastAPI","source_path":"apps/python-fastapi","version_fields":["python","fastapi","uvicorn","asyncpg"],"acceptance_test":"tests/test_python_fastapi_service.py","failure_test":"tests/test_python_fastapi_acceptance.py"},
-    {"id":"python-flask","language":"Python","framework":"Flask","display_name":"Python / Flask","source_path":"apps/python-flask","version_fields":["python","flask","gunicorn","psycopg","psycopg-binary","psycopg-pool"],"acceptance_test":"tests/test_python_flask_service.py","failure_test":"tests/test_python_flask_acceptance.py"},
+    {"id":"python-flask","language":"Python","framework":"Flask","display_name":"Python / Flask","source_path":"apps/python-flask","version_fields":["python","flask","psycopg","psycopg-binary","psycopg-pool"],"acceptance_test":"tests/test_python_flask_service.py","failure_test":"tests/test_python_flask_acceptance.py"},
     {"id":"java-spring-boot","language":"Java","framework":"Spring Boot","display_name":"Java / Spring Boot","source_path":"apps/java-spring-boot","version_fields":["java","spring-boot","tomcat","postgresql","hikaricp","jackson","gradle"],"acceptance_test":"tests/test_java_spring_boot_service.py","failure_test":"tests/test_java_spring_boot_acceptance.py"}
   ],
   "cohorts": {

From 1723648dd8f384174b55f3f7a825de8021ed2d85 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:09:15 +0900
Subject: [PATCH 23/52] chore: regenerate compatible registry projection

---
 site/registry.mjs | 1 -
 1 file changed, 1 deletion(-)

diff --git a/site/registry.mjs b/site/registry.mjs
index 027d943..24ae24a 100644
--- a/site/registry.mjs
+++ b/site/registry.mjs
@@ -145,7 +145,6 @@ export const REGISTRY = {
       "version_fields": [
         "python",
         "flask",
-        "gunicorn",
         "psycopg",
         "psycopg-binary",
         "psycopg-pool"

From 3cfcf9ae8d540e35f8e16a0d948fcb231105b0c5 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:10:38 +0900
Subject: [PATCH 24/52] test: confirm production Gunicorn profile without
 publishing

---
 .github/workflows/flask-gunicorn-tdd.yml | 30 ++++++++++++++++++++++++
 1 file changed, 30 insertions(+)

diff --git a/.github/workflows/flask-gunicorn-tdd.yml b/.github/workflows/flask-gunicorn-tdd.yml
index 22ad000..66cb407 100644
--- a/.github/workflows/flask-gunicorn-tdd.yml
+++ b/.github/workflows/flask-gunicorn-tdd.yml
@@ -19,3 +19,33 @@ jobs:
           python -m unittest discover -s tests -p 'test_benchmark_registry.py'
       - name: Real Flask container acceptance
         run: python tests/test_python_flask_service.py
+      - name: Non-publishing production-profile confirmation
+        run: |
+          python - <<'PY'
+          import json
+          from pathlib import Path
+          from benchmark.contract_test import run_contract
+          from benchmark.environment import DockerEnvironment
+          from benchmark.healthcheck import EXTERNAL_READINESS
+          from benchmark.install_oha import ensure_oha
+
+          out = Path(".cache/flask-gunicorn-confirmation")
+          env = DockerEnvironment(ensure_oha(), out, health_policy=EXTERNAL_READINESS)
+          try:
+              env.build("python-flask")
+              identity = env.start("python-flask")
+              checks = run_contract("http://127.0.0.1:8080", implementation="python-flask")
+              results = {
+                  endpoint: env.measure(endpoint, 10, 1)
+                  for endpoint in ("/json", "/db/42", "/cpu")
+              }
+              print("FLASK_GUNICORN_CONFIRMATION=" + json.dumps({
+                  "official": False,
+                  "publishable": False,
+                  "contract_checks": checks,
+                  "container": identity,
+                  "results": results,
+              }, sort_keys=True))
+          finally:
+              env.cleanup()
+          PY

From 0520e048b38c9218a3d46c88ee0a450ea93b1857 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:12:53 +0900
Subject: [PATCH 25/52] test: remove temporary Flask Gunicorn workflow

---
 .github/workflows/flask-gunicorn-tdd.yml | 51 ------------------------
 1 file changed, 51 deletions(-)
 delete mode 100644 .github/workflows/flask-gunicorn-tdd.yml

diff --git a/.github/workflows/flask-gunicorn-tdd.yml b/.github/workflows/flask-gunicorn-tdd.yml
deleted file mode 100644
index 66cb407..0000000
--- a/.github/workflows/flask-gunicorn-tdd.yml
+++ /dev/null
@@ -1,51 +0,0 @@
-name: Temporary Flask Gunicorn acceptance
-on:
-  push:
-    branches: [fix/flask-gunicorn-70]
-permissions:
-  contents: read
-jobs:
-  acceptance:
-    runs-on: ubuntu-24.04
-    timeout-minutes: 20
-    steps:
-      - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09
-      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
-        with:
-          python-version-file: apps/python-flask/.python-version
-      - name: Static and shared regression tests
-        run: |
-          python -m unittest tests.test_python_flask_acceptance tests.test_benchmark_environment
-          python -m unittest discover -s tests -p 'test_benchmark_registry.py'
-      - name: Real Flask container acceptance
-        run: python tests/test_python_flask_service.py
-      - name: Non-publishing production-profile confirmation
-        run: |
-          python - <<'PY'
-          import json
-          from pathlib import Path
-          from benchmark.contract_test import run_contract
-          from benchmark.environment import DockerEnvironment
-          from benchmark.healthcheck import EXTERNAL_READINESS
-          from benchmark.install_oha import ensure_oha
-
-          out = Path(".cache/flask-gunicorn-confirmation")
-          env = DockerEnvironment(ensure_oha(), out, health_policy=EXTERNAL_READINESS)
-          try:
-              env.build("python-flask")
-              identity = env.start("python-flask")
-              checks = run_contract("http://127.0.0.1:8080", implementation="python-flask")
-              results = {
-                  endpoint: env.measure(endpoint, 10, 1)
-                  for endpoint in ("/json", "/db/42", "/cpu")
-              }
-              print("FLASK_GUNICORN_CONFIRMATION=" + json.dumps({
-                  "official": False,
-                  "publishable": False,
-                  "contract_checks": checks,
-                  "container": identity,
-                  "results": results,
-              }, sort_keys=True))
-          finally:
-              env.cleanup()
-          PY

From 59a0728136d4110e65633a51311eecf2bddf67ca Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:14:54 +0900
Subject: [PATCH 26/52] test: update current Flask runtime metadata expectation

---
 tests/test_benchmark_java_candidate.py | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/tests/test_benchmark_java_candidate.py b/tests/test_benchmark_java_candidate.py
index ccd885e..d6c52a9 100644
--- a/tests/test_benchmark_java_candidate.py
+++ b/tests/test_benchmark_java_candidate.py
@@ -44,7 +44,7 @@
     "python-flask": {
         "python": "3.14.7",
         "flask": "3.1.3",
-        "waitress": "3.0.2",
+        "gunicorn": "26.0.0",
         "psycopg": "3.3.5",
         "psycopg-binary": "3.3.5",
         "psycopg-pool": "3.3.1",

From 469f1137b1539a49f8b1aade9512f681a6a0e9bd Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:14:56 +0900
Subject: [PATCH 27/52] test: update Flask candidate runtime contract

---
 tests/test_benchmark_python_flask_candidate.py | 13 +++++++------
 1 file changed, 7 insertions(+), 6 deletions(-)

diff --git a/tests/test_benchmark_python_flask_candidate.py b/tests/test_benchmark_python_flask_candidate.py
index d4bb9bf..1b7d019 100644
--- a/tests/test_benchmark_python_flask_candidate.py
+++ b/tests/test_benchmark_python_flask_candidate.py
@@ -9,17 +9,18 @@
 
 
 class PythonFlaskCandidateTests(unittest.TestCase):
-    def test_flask_profile_is_explicit_and_single_process(self):
+    def test_flask_profile_is_explicit_and_single_request_worker(self):
         self.assertEqual((APP / ".python-version").read_text().strip(), "3.14.7")
         requirements = set((APP / "requirements.in").read_text().splitlines())
         self.assertIn("flask==3.1.3", requirements)
-        self.assertIn("waitress==3.0.2", requirements)
+        self.assertIn("gunicorn==26.0.0", requirements)\n        self.assertNotIn("waitress==3.0.2", requirements)
         self.assertIn("psycopg==3.3.5", requirements)
         self.assertIn("psycopg-pool==3.3.1", requirements)
-        server = (APP / "benchmark_api" / "server.py").read_text()
-        self.assertIn("threads=1", server)
-        self.assertIn("create_server", server)
-        self.assertNotIn("gunicorn", server.lower())
+        dockerfile = (APP / "Dockerfile").read_text()
+        self.assertIn('"--workers", "1"', dockerfile)
+        self.assertIn('"--worker-class", "gthread"', dockerfile)
+        self.assertIn('"--threads", "1"', dockerfile)
+        self.assertIn('"--keep-alive", "5"', dockerfile)
 
     def test_flask_is_registered_without_mutating_legacy_cohort(self):
         registry = json.loads((ROOT / "benchmark" / "implementations.json").read_text())

From 905a64f97bbe57bfd5d17bed327832faa0f93bc0 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:17:32 +0900
Subject: [PATCH 28/52] test: fix Flask candidate assertion formatting

---
 tests/test_benchmark_python_flask_candidate.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/tests/test_benchmark_python_flask_candidate.py b/tests/test_benchmark_python_flask_candidate.py
index 1b7d019..8508aae 100644
--- a/tests/test_benchmark_python_flask_candidate.py
+++ b/tests/test_benchmark_python_flask_candidate.py
@@ -13,7 +13,8 @@ def test_flask_profile_is_explicit_and_single_request_worker(self):
         self.assertEqual((APP / ".python-version").read_text().strip(), "3.14.7")
         requirements = set((APP / "requirements.in").read_text().splitlines())
         self.assertIn("flask==3.1.3", requirements)
-        self.assertIn("gunicorn==26.0.0", requirements)\n        self.assertNotIn("waitress==3.0.2", requirements)
+        self.assertIn("gunicorn==26.0.0", requirements)
+        self.assertNotIn("waitress==3.0.2", requirements)
         self.assertIn("psycopg==3.3.5", requirements)
         self.assertIn("psycopg-pool==3.3.1", requirements)
         dockerfile = (APP / "Dockerfile").read_text()

From f6b3649231f0790e71660e455c877f989538a326 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:20:53 +0900
Subject: [PATCH 29/52] style: apply shared Ruff formatting

---
 benchmark/environment.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/benchmark/environment.py b/benchmark/environment.py
index 160e269..c688689 100644
--- a/benchmark/environment.py
+++ b/benchmark/environment.py
@@ -131,7 +131,8 @@ def normalized(arguments):
     service = state.get("Config", {}).get("Labels", {}).get("com.docker.compose.service")
     expected_servers = 2 if service == "python-flask" else 1
     require(
-        commands.count(server) == expected_servers and all(command in allowed for command in commands),
+        commands.count(server) == expected_servers
+        and all(command in allowed for command in commands),
         (
             f"expected {expected_servers} server process(es) and only independent health probes"
             if allow_health_probe

From f6fe1a569d4c6270d59de716c6a9febcf34880cc Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:20:56 +0900
Subject: [PATCH 30/52] style: apply shared Ruff formatting

---
 tests/test_benchmark_environment.py | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/tests/test_benchmark_environment.py b/tests/test_benchmark_environment.py
index 1129020..8c1aec0 100644
--- a/tests/test_benchmark_environment.py
+++ b/tests/test_benchmark_environment.py
@@ -170,9 +170,7 @@ def test_flask_gunicorn_allows_exactly_one_arbiter_and_one_worker(self):
         value["Path"] = "python"
         value["Args"] = ["-m", "benchmark_api.server"]
         output = (
-            "PID COMMAND\n"
-            "10 python -m benchmark_api.server\n"
-            "11 python -m benchmark_api.server\n"
+            "PID COMMAND\n10 python -m benchmark_api.server\n11 python -m benchmark_api.server\n"
         )
         environment.validate_processes(value, output, allow_health_probe=False)
 

From 56c539f85f8d90a03477a8791d03981d175588d1 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:25:23 +0900
Subject: [PATCH 31/52] fix: model historical Flask server version alternatives

---
 benchmark/implementations.json | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/benchmark/implementations.json b/benchmark/implementations.json
index bab355b..97afecb 100644
--- a/benchmark/implementations.json
+++ b/benchmark/implementations.json
@@ -9,7 +9,7 @@
     {"id":"node-fastify","language":"Node.js","framework":"Fastify","display_name":"Node.js / Fastify","source_path":"apps/node-fastify","version_fields":["node","fastify","pg"],"acceptance_test":"tests/test_node_fastify_service.py","failure_test":"tests/test_node_fastify_acceptance.py"},
     {"id":"node-express","language":"Node.js","framework":"Express","display_name":"Node.js / Express","source_path":"apps/node-express","version_fields":["node","express","pg"],"acceptance_test":"tests/test_node_express_service.py","failure_test":"tests/test_node_express_acceptance.py"},
     {"id":"python-fastapi","language":"Python","framework":"FastAPI","display_name":"Python / FastAPI","source_path":"apps/python-fastapi","version_fields":["python","fastapi","uvicorn","asyncpg"],"acceptance_test":"tests/test_python_fastapi_service.py","failure_test":"tests/test_python_fastapi_acceptance.py"},
-    {"id":"python-flask","language":"Python","framework":"Flask","display_name":"Python / Flask","source_path":"apps/python-flask","version_fields":["python","flask","psycopg","psycopg-binary","psycopg-pool"],"acceptance_test":"tests/test_python_flask_service.py","failure_test":"tests/test_python_flask_acceptance.py"},
+    {"id":"python-flask","language":"Python","framework":"Flask","display_name":"Python / Flask","source_path":"apps/python-flask","version_fields":["python","flask","psycopg","psycopg-binary","psycopg-pool"],"version_any_of":[["waitress","gunicorn"]],"acceptance_test":"tests/test_python_flask_service.py","failure_test":"tests/test_python_flask_acceptance.py"},
     {"id":"java-spring-boot","language":"Java","framework":"Spring Boot","display_name":"Java / Spring Boot","source_path":"apps/java-spring-boot","version_fields":["java","spring-boot","tomcat","postgresql","hikaricp","jackson","gradle"],"acceptance_test":"tests/test_java_spring_boot_service.py","failure_test":"tests/test_java_spring_boot_acceptance.py"}
   ],
   "cohorts": {

From c90a4f6130923f09454b355064e95ede4486898f Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:25:26 +0900
Subject: [PATCH 32/52] feat: validate alternative runtime version fields

---
 benchmark/registry.py | 51 ++++++++++++++++++++++++++++++++-----------
 1 file changed, 38 insertions(+), 13 deletions(-)

diff --git a/benchmark/registry.py b/benchmark/registry.py
index 8c8f6ce..607d026 100644
--- a/benchmark/registry.py
+++ b/benchmark/registry.py
@@ -29,19 +29,19 @@ def validate_registry(data: dict) -> dict:
     identifiers = []
     sources = []
     for spec in specs:
-        object_fields(
-            spec,
-            (
-                "id",
-                "language",
-                "framework",
-                "display_name",
-                "source_path",
-                "version_fields",
-                "acceptance_test",
-                "failure_test",
-            ),
-            "implementation registration",
+        required_spec_fields = {
+            "id",
+            "language",
+            "framework",
+            "display_name",
+            "source_path",
+            "version_fields",
+            "acceptance_test",
+            "failure_test",
+        }
+        require(
+            set(spec) in (required_spec_fields, required_spec_fields | {"version_any_of"}),
+            "unexpected implementation registration fields",
         )
         identifier = spec["id"]
         require(
@@ -78,6 +78,20 @@ def validate_registry(data: dict) -> dict:
             "invalid version field",
         )
         require(len(fields) == len(set(fields)), "duplicate version fields")
+        alternatives = spec.get("version_any_of", [])
+        require(type(alternatives) is list, "invalid version alternatives")
+        for group in alternatives:
+            require(
+                type(group) is list
+                and len(group) >= 2
+                and len(group) == len(set(group))
+                and all(
+                    type(key) is str and re.fullmatch(r"[a-z][a-z0-9_-]*", key)
+                    for key in group
+                )
+                and not (set(group) & set(fields)),
+                "invalid version alternative group",
+            )
     require(len(identifiers) == len(set(identifiers)), "duplicate implementation ID")
     require(len(sources) == len(set(sources)), "implementations must have independent source paths")
     cohorts = data["cohorts"]
@@ -123,6 +137,17 @@ def implementation(identifier: str) -> dict:
     raise BenchmarkFailure("unknown implementation ID")
 
 
+def validate_version_fields(identifier: str, values: dict) -> None:
+    spec = implementation(identifier)
+    require(type(values) is dict, "missing stack versions")
+    require(set(spec["version_fields"]) <= set(values), "missing stack versions")
+    for group in spec.get("version_any_of", []):
+        require(
+            sum(key in values for key in group) == 1,
+            "expected exactly one alternative stack version",
+        )
+
+
 def active_members() -> tuple[str, ...]:
     return tuple(REGISTRY["cohorts"][REGISTRY["active_cohort"]]["members"])
 

From 112a936366b122e44b446d03b2ce3cd629e40bd6 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:25:31 +0900
Subject: [PATCH 33/52] fix: validate current runtime version alternatives

---
 benchmark/environment.py | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/benchmark/environment.py b/benchmark/environment.py
index c688689..ea27025 100644
--- a/benchmark/environment.py
+++ b/benchmark/environment.py
@@ -224,10 +224,9 @@ def match(pattern, text):
         "version extractors must cover registered implementations",
     )
     for identifier, values in versions.items():
-        require(
-            set(implementation(identifier)["version_fields"]) <= set(values),
-            "missing required stack versions",
-        )
+        from .registry import validate_version_fields
+
+        validate_version_fields(identifier, values)
         require(
             all(
                 type(value) is str

From 93b2e15e76a00d7413767645649032a4dda2fca8 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:25:34 +0900
Subject: [PATCH 34/52] fix: validate historical and current Flask server
 provenance

---
 benchmark/report.py | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/benchmark/report.py b/benchmark/report.py
index dcafc7c..9ac3290 100644
--- a/benchmark/report.py
+++ b/benchmark/report.py
@@ -180,9 +180,8 @@ def validate_report(report: dict, *, expected_context: dict | None = None) -> No
     text(metadata.get("docker_compose"), "Docker Compose version required")
     versions = object_fields(metadata.get("versions"), members, "versions")
     for implementation in members:
-        keys = registry.implementation(implementation)["version_fields"]
         values = versions[implementation]
-        require(type(values) is dict and set(keys) <= set(values), "missing stack versions")
+        registry.validate_version_fields(implementation, values)
         for value in values.values():
             require(
                 type(value) is str and re.fullmatch(r"\d+\.\d+\.\d+", value) is not None,

From 5f69a8e76b3d7680eaa1a56f25a94dce7108cc08 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:25:39 +0900
Subject: [PATCH 35/52] refactor: reuse registry version validation

---
 benchmark/environment.py | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/benchmark/environment.py b/benchmark/environment.py
index ea27025..b2b6828 100644
--- a/benchmark/environment.py
+++ b/benchmark/environment.py
@@ -25,7 +25,7 @@
 from .java_versions import JAVA_VERSION
 from .java_versions import pinned_versions as java_pinned_versions
 from .process import ROOT, execute
-from .registry import active_members, implementation, implementation_ids
+from .registry import active_members, implementation, implementation_ids, validate_version_fields
 from .results import BenchmarkFailure, require, strict_json
 
 
@@ -224,8 +224,6 @@ def match(pattern, text):
         "version extractors must cover registered implementations",
     )
     for identifier, values in versions.items():
-        from .registry import validate_version_fields
-
         validate_version_fields(identifier, values)
         require(
             all(

From 49bd6fb12307aad0515825c740f17518333fe0e5 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:25:45 +0900
Subject: [PATCH 36/52] chore: regenerate runtime version alternatives

---
 site/registry.mjs | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/site/registry.mjs b/site/registry.mjs
index 24ae24a..4cb4ad8 100644
--- a/site/registry.mjs
+++ b/site/registry.mjs
@@ -142,6 +142,12 @@ export const REGISTRY = {
       "id": "python-flask",
       "language": "Python",
       "source_path": "apps/python-flask",
+      "version_any_of": [
+        [
+          "waitress",
+          "gunicorn"
+        ]
+      ],
       "version_fields": [
         "python",
         "flask",

From 90b7a01adc54aec5e962676e39947c924c498bfe Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 11:25:50 +0900
Subject: [PATCH 37/52] fix: validate alternative Flask server versions in
 Pages

---
 site/app.mjs | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/site/app.mjs b/site/app.mjs
index 7521ca2..16a00df 100644
--- a/site/app.mjs
+++ b/site/app.mjs
@@ -150,6 +150,10 @@ function validateReport(report) {
     for (const key of STACKS[id].version_fields) {
       assert(typeof versionSet[key] === "string" && /^\d+\.\d+\.\d+$/.test(versionSet[key]), `invalid ${id} ${key} version`);
     }
+    for (const group of STACKS[id].version_any_of ?? []) {
+      const present = group.filter((key) => Object.hasOwn(versionSet, key));
+      assert(present.length === 1, `expected exactly one ${id} alternative stack version`);
+    }
     assert(Object.values(versionSet).every((value) => typeof value === "string" && /^\d+\.\d+\.\d+$/.test(value)), "invalid stack version");
     assert(Array.isArray(backend.endpoints) && backend.endpoints.length === ENDPOINTS.length, "three endpoints required");
     backend.endpoints.forEach((entry, endpointIndex) => {

From 11c09df9c61fea697065ce992c1e07a3539706d7 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:44:40 +0900
Subject: [PATCH 38/52] style: format registry imports

---
 benchmark/environment.py | 7 ++++++-
 1 file changed, 6 insertions(+), 1 deletion(-)

diff --git a/benchmark/environment.py b/benchmark/environment.py
index b2b6828..93097d1 100644
--- a/benchmark/environment.py
+++ b/benchmark/environment.py
@@ -25,7 +25,12 @@
 from .java_versions import JAVA_VERSION
 from .java_versions import pinned_versions as java_pinned_versions
 from .process import ROOT, execute
-from .registry import active_members, implementation, implementation_ids, validate_version_fields
+from .registry import (
+    active_members,
+    implementation,
+    implementation_ids,
+    validate_version_fields,
+)
 from .results import BenchmarkFailure, require, strict_json
 
 

From f805bb7ed5db426d4e54e897eefd9d0ab10af59c Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:45:14 +0900
Subject: [PATCH 39/52] refactor: keep Flask server metadata compatibility
 simple

---
 benchmark/registry.py | 51 +++++++++++--------------------------------
 1 file changed, 13 insertions(+), 38 deletions(-)

diff --git a/benchmark/registry.py b/benchmark/registry.py
index 607d026..8c8f6ce 100644
--- a/benchmark/registry.py
+++ b/benchmark/registry.py
@@ -29,19 +29,19 @@ def validate_registry(data: dict) -> dict:
     identifiers = []
     sources = []
     for spec in specs:
-        required_spec_fields = {
-            "id",
-            "language",
-            "framework",
-            "display_name",
-            "source_path",
-            "version_fields",
-            "acceptance_test",
-            "failure_test",
-        }
-        require(
-            set(spec) in (required_spec_fields, required_spec_fields | {"version_any_of"}),
-            "unexpected implementation registration fields",
+        object_fields(
+            spec,
+            (
+                "id",
+                "language",
+                "framework",
+                "display_name",
+                "source_path",
+                "version_fields",
+                "acceptance_test",
+                "failure_test",
+            ),
+            "implementation registration",
         )
         identifier = spec["id"]
         require(
@@ -78,20 +78,6 @@ def validate_registry(data: dict) -> dict:
             "invalid version field",
         )
         require(len(fields) == len(set(fields)), "duplicate version fields")
-        alternatives = spec.get("version_any_of", [])
-        require(type(alternatives) is list, "invalid version alternatives")
-        for group in alternatives:
-            require(
-                type(group) is list
-                and len(group) >= 2
-                and len(group) == len(set(group))
-                and all(
-                    type(key) is str and re.fullmatch(r"[a-z][a-z0-9_-]*", key)
-                    for key in group
-                )
-                and not (set(group) & set(fields)),
-                "invalid version alternative group",
-            )
     require(len(identifiers) == len(set(identifiers)), "duplicate implementation ID")
     require(len(sources) == len(set(sources)), "implementations must have independent source paths")
     cohorts = data["cohorts"]
@@ -137,17 +123,6 @@ def implementation(identifier: str) -> dict:
     raise BenchmarkFailure("unknown implementation ID")
 
 
-def validate_version_fields(identifier: str, values: dict) -> None:
-    spec = implementation(identifier)
-    require(type(values) is dict, "missing stack versions")
-    require(set(spec["version_fields"]) <= set(values), "missing stack versions")
-    for group in spec.get("version_any_of", []):
-        require(
-            sum(key in values for key in group) == 1,
-            "expected exactly one alternative stack version",
-        )
-
-
 def active_members() -> tuple[str, ...]:
     return tuple(REGISTRY["cohorts"][REGISTRY["active_cohort"]]["members"])
 

From f80dfe8238ccea816fc8665f373f11f668073a5c Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:45:19 +0900
Subject: [PATCH 40/52] refactor: keep Flask server metadata compatibility
 simple

---
 benchmark/report.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/benchmark/report.py b/benchmark/report.py
index 9ac3290..dcafc7c 100644
--- a/benchmark/report.py
+++ b/benchmark/report.py
@@ -180,8 +180,9 @@ def validate_report(report: dict, *, expected_context: dict | None = None) -> No
     text(metadata.get("docker_compose"), "Docker Compose version required")
     versions = object_fields(metadata.get("versions"), members, "versions")
     for implementation in members:
+        keys = registry.implementation(implementation)["version_fields"]
         values = versions[implementation]
-        registry.validate_version_fields(implementation, values)
+        require(type(values) is dict and set(keys) <= set(values), "missing stack versions")
         for value in values.values():
             require(
                 type(value) is str and re.fullmatch(r"\d+\.\d+\.\d+", value) is not None,

From 3877701657d83aa76104ddacc9de2b859a8ae73d Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:45:26 +0900
Subject: [PATCH 41/52] refactor: keep Flask server metadata compatibility
 simple

---
 site/app.mjs | 4 ----
 1 file changed, 4 deletions(-)

diff --git a/site/app.mjs b/site/app.mjs
index 16a00df..7521ca2 100644
--- a/site/app.mjs
+++ b/site/app.mjs
@@ -150,10 +150,6 @@ function validateReport(report) {
     for (const key of STACKS[id].version_fields) {
       assert(typeof versionSet[key] === "string" && /^\d+\.\d+\.\d+$/.test(versionSet[key]), `invalid ${id} ${key} version`);
     }
-    for (const group of STACKS[id].version_any_of ?? []) {
-      const present = group.filter((key) => Object.hasOwn(versionSet, key));
-      assert(present.length === 1, `expected exactly one ${id} alternative stack version`);
-    }
     assert(Object.values(versionSet).every((value) => typeof value === "string" && /^\d+\.\d+\.\d+$/.test(value)), "invalid stack version");
     assert(Array.isArray(backend.endpoints) && backend.endpoints.length === ENDPOINTS.length, "three endpoints required");
     backend.endpoints.forEach((entry, endpointIndex) => {

From 89bbf56166dceaca067b4a19336d8db98131c670 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:45:29 +0900
Subject: [PATCH 42/52] refactor: make Flask server metadata additive

---
 benchmark/implementations.json | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/benchmark/implementations.json b/benchmark/implementations.json
index 97afecb..bab355b 100644
--- a/benchmark/implementations.json
+++ b/benchmark/implementations.json
@@ -9,7 +9,7 @@
     {"id":"node-fastify","language":"Node.js","framework":"Fastify","display_name":"Node.js / Fastify","source_path":"apps/node-fastify","version_fields":["node","fastify","pg"],"acceptance_test":"tests/test_node_fastify_service.py","failure_test":"tests/test_node_fastify_acceptance.py"},
     {"id":"node-express","language":"Node.js","framework":"Express","display_name":"Node.js / Express","source_path":"apps/node-express","version_fields":["node","express","pg"],"acceptance_test":"tests/test_node_express_service.py","failure_test":"tests/test_node_express_acceptance.py"},
     {"id":"python-fastapi","language":"Python","framework":"FastAPI","display_name":"Python / FastAPI","source_path":"apps/python-fastapi","version_fields":["python","fastapi","uvicorn","asyncpg"],"acceptance_test":"tests/test_python_fastapi_service.py","failure_test":"tests/test_python_fastapi_acceptance.py"},
-    {"id":"python-flask","language":"Python","framework":"Flask","display_name":"Python / Flask","source_path":"apps/python-flask","version_fields":["python","flask","psycopg","psycopg-binary","psycopg-pool"],"version_any_of":[["waitress","gunicorn"]],"acceptance_test":"tests/test_python_flask_service.py","failure_test":"tests/test_python_flask_acceptance.py"},
+    {"id":"python-flask","language":"Python","framework":"Flask","display_name":"Python / Flask","source_path":"apps/python-flask","version_fields":["python","flask","psycopg","psycopg-binary","psycopg-pool"],"acceptance_test":"tests/test_python_flask_service.py","failure_test":"tests/test_python_flask_acceptance.py"},
     {"id":"java-spring-boot","language":"Java","framework":"Spring Boot","display_name":"Java / Spring Boot","source_path":"apps/java-spring-boot","version_fields":["java","spring-boot","tomcat","postgresql","hikaricp","jackson","gradle"],"acceptance_test":"tests/test_java_spring_boot_service.py","failure_test":"tests/test_java_spring_boot_acceptance.py"}
   ],
   "cohorts": {

From c38ca3adc61e9167fe13af431a212d810331a86b Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:45:35 +0900
Subject: [PATCH 43/52] chore: regenerate Flask registry projection

---
 site/registry.mjs | 6 ------
 1 file changed, 6 deletions(-)

diff --git a/site/registry.mjs b/site/registry.mjs
index 4cb4ad8..24ae24a 100644
--- a/site/registry.mjs
+++ b/site/registry.mjs
@@ -142,12 +142,6 @@ export const REGISTRY = {
       "id": "python-flask",
       "language": "Python",
       "source_path": "apps/python-flask",
-      "version_any_of": [
-        [
-          "waitress",
-          "gunicorn"
-        ]
-      ],
       "version_fields": [
         "python",
         "flask",

From c32986c5b1b470199c3451cba1e3b0b6b0517273 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:46:06 +0900
Subject: [PATCH 44/52] refactor: scope Gunicorn process validation to Flask

---
 benchmark/environment.py | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/benchmark/environment.py b/benchmark/environment.py
index 93097d1..c688689 100644
--- a/benchmark/environment.py
+++ b/benchmark/environment.py
@@ -25,12 +25,7 @@
 from .java_versions import JAVA_VERSION
 from .java_versions import pinned_versions as java_pinned_versions
 from .process import ROOT, execute
-from .registry import (
-    active_members,
-    implementation,
-    implementation_ids,
-    validate_version_fields,
-)
+from .registry import active_members, implementation, implementation_ids
 from .results import BenchmarkFailure, require, strict_json
 
 
@@ -229,7 +224,10 @@ def match(pattern, text):
         "version extractors must cover registered implementations",
     )
     for identifier, values in versions.items():
-        validate_version_fields(identifier, values)
+        require(
+            set(implementation(identifier)["version_fields"]) <= set(values),
+            "missing required stack versions",
+        )
         require(
             all(
                 type(value) is str

From b611e7cf3e64516ea95efce9bcc7dccea83bdcba Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:49:43 +0900
Subject: [PATCH 45/52] test: capture site compatibility failure

---
 .github/workflows/site-test-diagnostic.yml | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)
 create mode 100644 .github/workflows/site-test-diagnostic.yml

diff --git a/.github/workflows/site-test-diagnostic.yml b/.github/workflows/site-test-diagnostic.yml
new file mode 100644
index 0000000..36e5d36
--- /dev/null
+++ b/.github/workflows/site-test-diagnostic.yml
@@ -0,0 +1,22 @@
+name: Temporary site failure diagnostic
+on:
+  push:
+    branches: [fix/flask-gunicorn-70]
+permissions:
+  contents: read
+jobs:
+  diagnose:
+    runs-on: ubuntu-24.04
+    steps:
+      - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09
+      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
+        with:
+          python-version: '3.10'
+      - run: |
+          mkdir -p .cache
+          python -m unittest discover -s tests -p 'test_site.py' -v > .cache/site-test.log 2>&1 || true
+      - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
+        with:
+          name: site-test-diagnostic
+          path: .cache/site-test.log
+          retention-days: 1

From b662ac556f8ff192fc712257c1d094261cf15e93 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:50:24 +0900
Subject: [PATCH 46/52] test: reproduce site failure on CI Python

---
 .github/workflows/site-test-diagnostic.yml | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/.github/workflows/site-test-diagnostic.yml b/.github/workflows/site-test-diagnostic.yml
index 36e5d36..4b6f74b 100644
--- a/.github/workflows/site-test-diagnostic.yml
+++ b/.github/workflows/site-test-diagnostic.yml
@@ -11,7 +11,7 @@ jobs:
       - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09
       - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
         with:
-          python-version: '3.10'
+          python-version-file: apps/python-fastapi/.python-version
       - run: |
           mkdir -p .cache
           python -m unittest discover -s tests -p 'test_site.py' -v > .cache/site-test.log 2>&1 || true

From eb7007706b94931c921400ccdaa2b54621a4e047 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:51:04 +0900
Subject: [PATCH 47/52] test: remove temporary site diagnostic

---
 .github/workflows/site-test-diagnostic.yml | 22 ----------------------
 1 file changed, 22 deletions(-)
 delete mode 100644 .github/workflows/site-test-diagnostic.yml

diff --git a/.github/workflows/site-test-diagnostic.yml b/.github/workflows/site-test-diagnostic.yml
deleted file mode 100644
index 4b6f74b..0000000
--- a/.github/workflows/site-test-diagnostic.yml
+++ /dev/null
@@ -1,22 +0,0 @@
-name: Temporary site failure diagnostic
-on:
-  push:
-    branches: [fix/flask-gunicorn-70]
-permissions:
-  contents: read
-jobs:
-  diagnose:
-    runs-on: ubuntu-24.04
-    steps:
-      - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09
-      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
-        with:
-          python-version-file: apps/python-fastapi/.python-version
-      - run: |
-          mkdir -p .cache
-          python -m unittest discover -s tests -p 'test_site.py' -v > .cache/site-test.log 2>&1 || true
-      - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
-        with:
-          name: site-test-diagnostic
-          path: .cache/site-test.log
-          retention-days: 1

From 70cdbbba75b9b75dec131316fef675d406b995af Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:53:03 +0900
Subject: [PATCH 48/52] test: reproduce shared site failure order

---
 .github/workflows/shared-order-diagnostic.yml | 28 +++++++++++++++++++
 1 file changed, 28 insertions(+)
 create mode 100644 .github/workflows/shared-order-diagnostic.yml

diff --git a/.github/workflows/shared-order-diagnostic.yml b/.github/workflows/shared-order-diagnostic.yml
new file mode 100644
index 0000000..6f7e124
--- /dev/null
+++ b/.github/workflows/shared-order-diagnostic.yml
@@ -0,0 +1,28 @@
+name: Temporary shared-order diagnostic
+on:
+  push:
+    branches: [fix/flask-gunicorn-70]
+permissions:
+  contents: read
+jobs:
+  diagnose:
+    runs-on: ubuntu-24.04
+    steps:
+      - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09
+      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
+        with:
+          python-version: '3.10'
+      - run: make test-benchmark
+      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
+        with:
+          python-version-file: apps/python-fastapi/.python-version
+      - run: python -m pip install --require-hashes --only-binary=:all: -r apps/python-fastapi/requirements-dev.lock
+      - run: make test-registry
+      - run: |
+          mkdir -p .cache/diag
+          python -m unittest discover -s tests -p 'test_site.py' -v > .cache/diag/site.log 2>&1 || true
+      - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
+        with:
+          name: shared-order-diagnostic
+          path: .cache/diag/site.log
+          retention-days: 1

From 9c891d6495054a102e6a03f619f1ed1812b514c4 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:53:48 +0900
Subject: [PATCH 49/52] test: capture shared site failure output

---
 .github/workflows/ci.yml | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 9c8610a..269195e 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -84,7 +84,7 @@ jobs:
       - name: make test-registry
         run: make test-registry
       - name: make test-site
-        run: make test-site
+        run: make test-site > .cache/ci/site.log 2>&1
       - name: make test-db
         run: make test-db
       - name: make test-benchmark
@@ -99,7 +99,9 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
         with:
           name: formatting-${{ github.run_id }}-${{ github.run_attempt }}
-          path: .cache/ci/format.patch
+          path: |
+            .cache/ci/format.patch
+            .cache/ci/site.log
           include-hidden-files: true
           retention-days: 7
           if-no-files-found: ignore

From 511032bf8067b3aa9851b7697f909713697489ac Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:53:53 +0900
Subject: [PATCH 50/52] test: remove temporary shared-order diagnostic

---
 .github/workflows/shared-order-diagnostic.yml | 28 -------------------
 1 file changed, 28 deletions(-)
 delete mode 100644 .github/workflows/shared-order-diagnostic.yml

diff --git a/.github/workflows/shared-order-diagnostic.yml b/.github/workflows/shared-order-diagnostic.yml
deleted file mode 100644
index 6f7e124..0000000
--- a/.github/workflows/shared-order-diagnostic.yml
+++ /dev/null
@@ -1,28 +0,0 @@
-name: Temporary shared-order diagnostic
-on:
-  push:
-    branches: [fix/flask-gunicorn-70]
-permissions:
-  contents: read
-jobs:
-  diagnose:
-    runs-on: ubuntu-24.04
-    steps:
-      - uses: actions/checkout@fbc6f3992d24b796d5a048ff273f7fcc4a7b6c09
-      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
-        with:
-          python-version: '3.10'
-      - run: make test-benchmark
-      - uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065
-        with:
-          python-version-file: apps/python-fastapi/.python-version
-      - run: python -m pip install --require-hashes --only-binary=:all: -r apps/python-fastapi/requirements-dev.lock
-      - run: make test-registry
-      - run: |
-          mkdir -p .cache/diag
-          python -m unittest discover -s tests -p 'test_site.py' -v > .cache/diag/site.log 2>&1 || true
-      - uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
-        with:
-          name: shared-order-diagnostic
-          path: .cache/diag/site.log
-          retention-days: 1

From 8e162e376c2eac225901aa720c51c48eda687fd2 Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:58:03 +0900
Subject: [PATCH 51/52] test: validate required stack version fields from
 registry

---
 tests/eight_stack_cases.mjs | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tests/eight_stack_cases.mjs b/tests/eight_stack_cases.mjs
index a1db6ee..0362f91 100644
--- a/tests/eight_stack_cases.mjs
+++ b/tests/eight_stack_cases.mjs
@@ -6,6 +6,8 @@ import { join } from 'node:path';
 import { fileURLToPath } from 'node:url';
 import test from 'node:test';
 
+import { REGISTRY } from '../site/registry.mjs';
+
 const members = ['go-gin', 'go-echo', 'rust-actix', 'rust-axum', 'node-fastify', 'node-express', 'python-fastapi', 'python-flask'];
 const legacyMembers = ['go-gin', 'rust-actix', 'node-fastify', 'python-fastapi'];
 
@@ -91,7 +93,8 @@ export function registerEightStackTests() {
         r => { r.implementations[index].endpoints.reverse(); },
         r => { r.implementations[index].endpoints[0].runs.pop(); },
       );
-      for (const field of Object.keys(eight.metadata.versions[members[index]])) {
+      const spec = REGISTRY.implementations.find(item => item.id === members[index]);
+      for (const field of spec.version_fields) {
         changes.push(r => { delete r.metadata.versions[members[index]][field]; });
       }
     }

From 75b108fc475672554c249daeb29f25ea8d56e9dd Mon Sep 17 00:00:00 2001
From: Yuma Takino <36617106+tappe9@users.noreply.github.com>
Date: Sun, 27 Sep 2026 18:58:09 +0900
Subject: [PATCH 52/52] test: remove temporary CI failure capture

---
 .github/workflows/ci.yml | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 269195e..9c8610a 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -84,7 +84,7 @@ jobs:
       - name: make test-registry
         run: make test-registry
       - name: make test-site
-        run: make test-site > .cache/ci/site.log 2>&1
+        run: make test-site
       - name: make test-db
         run: make test-db
       - name: make test-benchmark
@@ -99,9 +99,7 @@ jobs:
         uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02
         with:
           name: formatting-${{ github.run_id }}-${{ github.run_attempt }}
-          path: |
-            .cache/ci/format.patch
-            .cache/ci/site.log
+          path: .cache/ci/format.patch
           include-hidden-files: true
           retention-days: 7
           if-no-files-found: ignore