Probe Fuseki by TCP instead of /$/ping

On the test server Fuseki 5.1's Java heap grew by about 1.2 GiB per hour
from the moment the monitoring started, from a flat ~220 MiB before.
Stopping only the probes for four minutes took the growth from 20 to
1.5 MiB per minute, so the HTTP ping every 15 s is the cause. A TCP
connect to port 3030 still shows whether Fuseki is up, without making
the JVM do anything; a hung JVM with an open port is no longer caught.

Part of #2412.

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

Author: jh-RLI

podman/monitoring/prometheus/prometheus.yml

@@ -35,14 +35,16 @@ scrape_configs:
         target_label: instance
       - target_label: __address__
         replacement: oep-mon-blackbox:9115
-  - job_name: probe_fuseki
-    metrics_path: /probe
-    params: { module: [http_2xx] }
-    static_configs: [{ targets: ["http://fuseki:3030/$/ping"] }]
-    relabel_configs: *probe
   - job_name: probe_tcp
     metrics_path: /probe
     params: { module: [tcp_connect] }
+    # Fuseki by TCP only: an HTTP probe of /$/ping every 15 s made Fuseki 5.1's
+    # Java heap grow by ~1.2 GiB per hour (measured on the test server, 2026-10-14).
     static_configs:
-      [{ targets: ["ontop:8080", "lookup:8082", "postgres:5432"] }]
+      [
+        {
+          targets:
+            ["fuseki:3030", "ontop:8080", "lookup:8082", "postgres:5432"],
+        },
+      ]
     relabel_configs: *probe