feat(openzaak): one-command UI-OpenZaak bridge, real BSN fix, flake mitigation

scripts/openzaak-ui-up.sh brings up the root app and the OpenZaak harness
together, wires them onto one docker network, seeds the catalogus, grants
the container-alias zaaktype scope, and verifies a real aanvraag submitted
through the UI lands in OpenZaak.

Along the way: DocumentStore.DemoOwner was reusing the seeded doctor's
11-digit BIG-nummer as a stand-in BSN, which isn't a valid 9-digit BSN shape
— OpenZaak rejects it, breaking both submit's rol-creation step and the
citizen's own applications list under Zgw:Enabled=true. Fixed to a real
elfproef-valid BSN.

Also adds mitigation for a still-unexplained per-container flake (every
outbound ZGW POST fails as if the body were empty, for that container's
whole lifetime) that correlates with host memory pressure: the script now
warns when host swap is heavily used, and an opt-in ZgwDiagnosticHandler
(ZGW_DEBUG_HTTP=1) logs Content-Length vs. actual bytes sent so the next
reproduction can confirm or rule out client-side body corruption.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

scripts/openzaak-ui-up.sh

@@ -0,0 +1,150 @@
+#!/usr/bin/env bash
+# One command to test the Angular UI end-to-end against a real OpenZaak instead of the local
+# SQLite store: brings up the OpenZaak harness (backend/openzaak/), seeds its catalogus, wires
+# it onto the root app's docker network (docker-compose.openzaak.bff.yml /
+# docker-compose.openzaak.yml — see those files for the full "why", it's non-obvious), grants
+# the extra authorization scope the container alias needs, and brings the FE+BFF up pointed at
+# OpenZaak.
+#
+# Safe to re-run: every step this chains is already idempotent (bootstrap-catalogus.sh,
+# `docker compose up -d`, and the grant-replace below).
+#
+# Also confirmed empirically, many repeated trials: some freshly-(re)started `api`
+# containers have EVERY outbound ZGW POST fail with what looks like an empty body reaching
+# OpenZaak ("all fields required"), for that container's entire lifetime — while a plain curl
+# to the exact same URL, from inside the exact same container, never fails, even hammered in a
+# loop. Ruled out as the cause: HttpClient connection pooling settings, Expect-100-Continue,
+# content pre-buffering — none of it made a measurable difference. Best lead so far, and
+# reproduced live on this dev host (7.5/8GB swap from long-idle unrelated containers): this
+# correlates with the HOST being under heavy memory pressure — a heavier managed runtime
+# (dotnet's JIT + GC) is plausibly far more sensitive to that than a lightweight one-shot
+# `curl` process. If you hit this, try freeing host memory (stop unrelated containers) before
+# assuming it's a code regression. The preflight check and self-check below warn about and
+# retry around it either way; set ZGW_DEBUG_HTTP=1 on the `api` container (see
+# docker-compose.openzaak.yml) next time it reproduces to log Content-Length vs. actual bytes
+# sent, which would confirm (or rule out) client-side body corruption.
+set -euo pipefail
+cd "$(dirname "$0")/.."
+
+step() { printf '\n\033[1;36m▶ %s\033[0m\n' "$1"; }
+
+check_memory_pressure() {
+  local total used pct
+  read -r total used <<< "$(free -m | awk '/^Swap:/{print $2, $3}')"
+  [ "${total:-0}" -gt 0 ] || return 0
+  pct=$(( used * 100 / total ))
+  if [ "$pct" -ge 50 ]; then
+    echo "⚠ host swap ${pct}% used (${used}MiB/${total}MiB) — this correlates with the" >&2
+    echo "  known per-container ZGW flake (see comment up top). Consider 'docker ps' and" >&2
+    echo "  stopping unrelated long-running stacks before continuing." >&2
+  fi
+}
+check_memory_pressure
+
+step "root app (creates the docker network the OpenZaak harness joins below)"
+docker compose up -d
+
+step "OpenZaak harness + bff overlay"
+( cd backend/openzaak && docker compose -f docker-compose.openzaak.yml -f docker-compose.openzaak.bff.yml up -d )
+
+step "seed catalogus/zaaktype/zaak (idempotent)"
+bootstrap_output=$(cd backend/openzaak && ./bootstrap-catalogus.sh)
+echo "$bootstrap_output"
+
+zaaktype_line=$(printf '%s\n' "$bootstrap_output" | grep -A1 '^Zaaktype (concept)\.\.\.$' | tail -n1)
+zaaktype_url=$(printf '%s\n' "$zaaktype_line" | sed -E 's/^ *(exists|created): //')
+zaaktype_uuid=$(printf '%s\n' "$zaaktype_url" | sed 's#.*/##')
+if [ -z "$zaaktype_uuid" ]; then
+  echo "Could not find the seeded zaaktype's URL in bootstrap-catalogus.sh's output — see above." >&2
+  exit 1
+fi
+# bootstrap-catalogus.sh queried (and granted scope) via http://localhost:8000 (run from the
+# host); the containerized BFF reaches the same resource via the `openzaak.local` alias
+# (docker-compose.openzaak.bff.yml) instead — only the UUID suffix is what actually matters.
+container_zaaktype_url="http://openzaak.local:8000/catalogi/api/v1/zaaktypen/${zaaktype_uuid}"
+export OPENZAAK_ZAAKTYPE_URL="$container_zaaktype_url"
+
+step "grant the container-alias zaaktype scope (REPLACES bootstrap-catalogus.sh's own localhost-scoped grant, doesn't add to it — OpenZaak's own zaken-list authorization filter 500s with a RuntimeError, 'are you sure that all paths point to the same resource?', when an Applicatie has two zrc grants for what's really the same zaaktype under two different hostnames; confirmed empirically. One consequence: dotnet test --filter Category=Integration needs bootstrap-catalogus.sh rerun afterward to restore its localhost-scoped grant — it's idempotent, so that's a plain rerun, not a reset)"
+( cd backend/openzaak && docker compose -f docker-compose.openzaak.yml exec -T --workdir /app/src web python manage.py shell ) <<PY
+from vng_api_common.authorizations.models import Applicatie
+app = Applicatie.objects.get(client_ids__contains=["bigregister-test"])
+app.autorisaties.filter(component="zrc").delete()
+app.autorisaties.create(
+    component="zrc",
+    scopes=["zaken.aanmaken", "zaken.bijwerken", "zaken.lezen"],
+    zaaktype="$container_zaaktype_url",
+    max_vertrouwelijkheidaanduiding="openbaar",
+)
+print("granted:", "$container_zaaktype_url")
+PY
+
+step "root app again (now pointed at OpenZaak)"
+docker compose -f docker-compose.yml -f docker-compose.openzaak.yml up -d
+
+step "verify the BFF can actually reach OpenZaak (retries api if not — see the flake note up top)"
+wait_for_api() {
+  # -f: a response has to actually be 2xx — dotnet run's own build/start (and, on a retry, a
+  # full restart) can leave the port accepting-but-erroring for a few seconds first, which a
+  # plain curl (no -f) would misread as "ready".
+  for _ in $(seq 1 60); do
+    curl -sS -f -m 2 -o /dev/null http://localhost:5000/api/v1/me && return 0
+    sleep 1
+  done
+  return 1
+}
+# A throwaway herregistratie submit: does the resulting zaak actually show up in OpenZaak
+# (via the BFF's own /admin/cases)?
+zgw_write_reaches_openzaak() {
+  local create id ref cases
+  create=$(curl -sS -m 10 -X POST http://localhost:5000/api/v1/applications -H "Content-Type: application/json" -d '{"type":"herregistratie"}') || return 1
+  id=$(printf '%s' "$create" | python3 -c 'import json,sys; print(json.load(sys.stdin)["id"])') || return 1
+  ref=$(curl -sS -m 15 -X POST "http://localhost:5000/api/v1/applications/$id/submit" -H "Content-Type: application/json" -d '{"uren":1}' | python3 -c 'import json,sys; print(json.load(sys.stdin)["referentie"])') || return 1
+  cases=$(curl -sS -m 10 -H "X-Role: admin" http://localhost:5000/api/v1/admin/cases)
+  printf '%s' "$cases" | python3 -c "
+import json, sys
+refs = [c['status']['referentie'] for c in json.load(sys.stdin)]
+sys.exit(0 if '$ref' in refs else 1)
+"
+}
+verified=false
+attempts=5
+for attempt in $(seq 1 "$attempts"); do
+  wait_for_api || { echo "api never came up listening — see 'docker logs atomic-design-poc-api-1'." >&2; break; }
+  if zgw_write_reaches_openzaak; then
+    verified=true
+    break
+  fi
+  echo "  attempt $attempt/$attempts: a write didn't reach OpenZaak — restarting api and giving the network a moment to settle (known flake, see comment up top)..."
+  sleep 3
+  docker compose -f docker-compose.yml -f docker-compose.openzaak.yml restart api
+done
+if [ "$verified" = true ]; then
+  echo "  verified: a write reaches OpenZaak."
+else
+  check_memory_pressure
+  echo "  Could not verify after $attempts tries. This is the known flake, not necessarily a real failure —" >&2
+  echo "  keep retrying by hand: 'docker compose -f docker-compose.yml -f docker-compose.openzaak.yml restart api'," >&2
+  echo "  wait for 'Application started' in 'docker logs -f atomic-design-poc-api-1', then try the UI again." >&2
+fi
+
+printf '\n\033[1;32m✔ up\033[0m\n'
+cat <<EOF
+
+App:               http://localhost:4200
+Admin cases (proof an aanvraag landed in OpenZaak): http://localhost:4200/beheer/zaken?role=admin
+Audit trail (should show no zgw:divergence rows):   http://localhost:4200/beheer/audit?role=admin
+OpenZaak directly:  http://localhost:8000
+
+Only "herregistratie" has a seeded zaaktype in this harness — submit that wizard to see a real
+write land in OpenZaak. registratie/intake submissions still succeed locally but their ZGW
+write is flagged (Aanvraag.ZgwError / a zgw:divergence audit row), not surfaced as a UI error.
+Document uploads also won't register in ZGW (no Documenten content seeded) — pick "per post"
+in the wizard's document step to keep the demo clean, or ignore it, it's non-fatal.
+
+Tear down (plain "down", no -f overlay needed — overlay files only add env/network config to
+services docker-compose.yml already defines, and the required-var syntax in
+docker-compose.openzaak.yml would otherwise block "down" too once OPENZAAK_ZAAKTYPE_URL falls
+out of your shell):
+  docker compose down
+  ( cd backend/openzaak && docker compose -f docker-compose.openzaak.yml down -v )
+EOF