Health check builds a new Elasticsearch client on every request

api/views/health.py constructs a new Elasticsearch client on every call:

es_settings = settings.ELASTICSEARCH_DSL["default"]
es = Elasticsearch(hosts=es_settings["hosts"], http_auth=es_settings["http_auth"])
if es.ping():

Visible in the logs as a new TCP connection each time:

{"event": "Starting new HTTP connection (1): elasticsearch:9200"}
{"event": "http://elasticsearch:9200 \"HEAD / HTTP/1.1\" 200 0"}

Why it is worth fixing

/health/ is called by the Docker healthcheck, the deploy gate in scripts/ci-deploy.sh, and any external monitoring — so this is a fresh connection several times a minute, forever, to send one HEAD /.

Measured cost is small (the ping itself is ~0.14s, and the endpoint answers in ~0.04s on-box), so this is efficiency and log noise rather than a performance problem.

It also emits a DeprecationWarning on every call:

The 'http_auth' parameter is deprecated. Use 'basic_auth' or 'bearer_auth' parameters instead

Suggested

• Reuse a module-level client, or the one django-elasticsearch-dsl already configures, instead of building one per request.
• Move http_auth to basic_auth while there.

Worth keeping the check itself: it is what makes /health/ return 503 when Elasticsearch is down, which the deploy gate depends on. Only the per-request construction should change.

Low priority. Related: the health check's database probe was fixed in #140 to open a fresh connection deliberately — that one is intentional and should stay.