Replace non-functional SEQUENTIAL config with RANDOM ID reproduction test

Remove injectSysConfig("id.generation.strategy", "SEQUENTIAL") which did not
actually trigger the flake. Add KafkaClientDsmDisabledRandomIdsForkedTest that
overrides idGenerationStrategyName() to "RANDOM", matching production behavior.

With RANDOM IDs, SORT_TRACES_BY_ID produces non-deterministic order, causing
the original positional consumer-to-producer mapping to fail ~95% of the time.
Switch the batch consume test to SORT_TRACES_BY_START so the parent trace
(started before any consumer receives messages) is always at index 0. The
dynamic parent lookup fix handles any ordering of the 3 consumer traces.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: bm1549

dd-java-agent/instrumentation/kafka/kafka-clients-0.11/src/test/groovy/KafkaClientTestBase.groovy

@@ -898,7 +898,10 @@ abstract class KafkaClientTestBase extends VersionedNamingTestBase {
     }
     assert receivedSet.isEmpty()
 
-    assertTraces(4, SORT_TRACES_BY_ID) {
+    // Use SORT_TRACES_BY_START so the parent trace (started first, before any consumer
+    // receives messages) is always at index 0 regardless of span ID generation strategy.
+    // The dynamic parent lookup below handles any ordering of the 3 consumer traces.
+    assertTraces(4, SORT_TRACES_BY_START) {
       trace(7) {
         basicSpan(it, "parent")
         basicSpan(it, "producer callback", span(0))
@@ -923,7 +926,7 @@ abstract class KafkaClientTestBase extends VersionedNamingTestBase {
           queueSpan(it, trace(0)[2])
         }
       } else {
-        // Consumer traces are sorted by random span ID, so we can't assume a fixed
+        // Consumer traces are sorted by start time, so we can't assume a fixed
         // mapping between consumer trace index and producer span index. Instead, find
         // the actual parent producer span for each consumer trace dynamically.
         def producerSpans = [trace(0)[2], trace(0)[4], trace(0)[6]]
@@ -1481,10 +1484,6 @@ class KafkaClientDataStreamsDisabledForkedTest extends KafkaClientTestBase {
     super.configurePreAgent()
     injectSysConfig("dd.service", "KafkaClientDataStreamsDisabledForkedTest")
     injectSysConfig("dd.kafka.legacy.tracing.enabled", "true")
-    // Deterministic reproduction: SEQUENTIAL ID strategy forces a known sort order
-    // that differs from the reverse order the original positional code assumed,
-    // proving the dynamic parent lookup fix handles any ordering.
-    injectSysConfig("id.generation.strategy", "SEQUENTIAL")
   }
 
   @Override
@@ -1515,3 +1514,24 @@ class KafkaClientContextSwapForkedTest extends KafkaClientV0ForkedTest {
     injectSysConfig(TraceInstrumentationConfig.LEGACY_CONTEXT_MANAGER_ENABLED, "false")
   }
 }
+
+/**
+ * Reproduces the flake in "test spring kafka template produce and batch consume"
+ * by using RANDOM IDs (instead of the default SEQUENTIAL used in tests).
+ *
+ * Root cause: The test's assertTraces(4, SORT_TRACES_BY_ID) sorts traces by
+ * localRootSpan.spanId, then hardcodes positional mappings between consumer and
+ * producer traces. With SEQUENTIAL IDs (the test default), both the producer span
+ * finish order within trace(0) and the consumer trace sort order are driven by the
+ * same Kafka internal ordering, so the mapping happens to be consistent.
+ *
+ * With RANDOM IDs (as used in production), the sort order becomes non-deterministic.
+ * There are 3! = 6 possible orderings for the 3 consumer traces, and only 1 matches
+ * the hardcoded mapping. The dynamic parent lookup fix handles any ordering.
+ */
+class KafkaClientDsmDisabledRandomIdsForkedTest extends KafkaClientDataStreamsDisabledForkedTest {
+  @Override
+  protected String idGenerationStrategyName() {
+    return "RANDOM"
+  }
+}