Stabilize BatchDriver single-batch test startup

muted-tests.yml

@@ -372,9 +372,9 @@ tests:
 - class: org.elasticsearch.xpack.search.AsyncSearchTaskTests
   method: testWithFailure
   issue: https://github.com/elastic/elasticsearch/issues/142821
-- class: org.elasticsearch.compute.operator.exchange.BatchDriverTests
-  method: testSinglePageSingleBatch
-  issue: https://github.com/elastic/elasticsearch/issues/142895
+- class: org.elasticsearch.xpack.esql.qa.multi_node.EsqlClientYamlIT
+  method: test {p0=esql/40_tsdb/to_aggregate_metric_double with multi_values}
+  issue: https://github.com/elastic/elasticsearch/issues/142964
 - class: org.elasticsearch.reindex.management.ReindexManagementClientYamlTestSuiteIT
   method: test {yaml=reindex/30_cancel_reindex/Cancel running reindex returns response and GET confirms completed}
   issue: https://github.com/elastic/elasticsearch/issues/142079

x-pack/plugin/esql/compute/src/test/java/org/elasticsearch/compute/operator/exchange/BatchDriverTests.java

@@ -804,8 +804,9 @@ private void runDriverAndWait(ThreadPool threadPool, BatchDriver batchDriver, Ru
         // This helps catch timing-related issues that might be hidden when everything runs sequentially
         Thread batchFeedingThread = new Thread(() -> {
             try {
-                // Feed first batch - this will trigger the callback chain
-                threadPool.executor(ThreadPool.Names.SEARCH).execute(feedBatch);
+                // Feed first batch directly to avoid racing with driver scheduling on the same SEARCH executor.
+                // Subsequent batches are still triggered by onBatchEnd callbacks.
+                feedBatch.run();
             } catch (Exception e) {
                 logger.error("[TEST] Error in batch feeding thread", e);
                 throw new AssertionError("Error in batch feeding thread", e);