diff --git a/amber/src/test/scala/org/apache/texera/web/resource/dashboard/DatasetSearchQueryBuilderSpec.scala b/amber/src/test/scala/org/apache/texera/web/resource/dashboard/DatasetSearchQueryBuilderSpec.scala
index 6987540fac1..6be27bc25ce 100644
--- a/amber/src/test/scala/org/apache/texera/web/resource/dashboard/DatasetSearchQueryBuilderSpec.scala
+++ b/amber/src/test/scala/org/apache/texera/web/resource/dashboard/DatasetSearchQueryBuilderSpec.scala
@@ -111,16 +111,21 @@ import scala.jdk.CollectionConverters._
   * `WorkflowExecutionService` a test->test dependency on `DAO` and `Auth` only, so workflow-core's
   * test tree is not on this module's test classpath.
   *
-  * The keyword tests RENDER a full-text predicate (they do not fetch one), which does read the
-  * JVM-global `FulltextSearchQueryUtils.usePgroonga` and emits whichever arm it currently selects:
-  * `pgroonga_condition(...)` when this suite runs alone, the `to_tsvector`/`to_tsquery` arm if
-  * `DatasetResourceSpec` or `WorkflowResourceSpec` ran earlier in this JVM and left the global
-  * `false` (both set it and neither restores it; amber has no `Test / fork`). This suite therefore
-  * neither touches nor restores that global, and every keyword assertion here is deliberately
-  * branch-independent: the tokens themselves and the `coalesce(...) || ' ' || coalesce(...)`
-  * expression are built at `FulltextSearchQueryUtils:49-51`, *before* the `if (usePgroonga)`.
-  * Anything added here must keep that property — an assertion on `pgroonga_condition` would pass
-  * solo and fail in a full-module run.
+  * The keyword tests RENDER a full-text predicate (they do not fetch one), and rendering reads the
+  * JVM-global `FulltextSearchQueryUtils.usePgroonga`: the `pgroonga_condition(...)` arm while it
+  * holds `true`, the `to_tsvector`/`to_tsquery` arm while it holds `false`. That flag is a plain
+  * mutable `var` and amber has no `Test / fork`, so its value here is whatever the suites sharing
+  * this JVM have left it at — not something this spec controls or should assume. This suite
+  * therefore neither touches nor restores it, and every keyword assertion here is deliberately
+  * branch-independent. Two things reach both arms: the `coalesce(...) || ' ' || coalesce(...)`
+  * expression, built in `FulltextSearchQueryUtils` as `combinedFields` before the `if (usePgroonga)`
+  * branch and embedded verbatim by either arm, and each INDIVIDUAL keyword token. Their JOINING does
+  * not — the `true` arm space-joins the full keyword list into one literal (rendering
+  * `pgroonga_condition('alpha beta', ...)`), while the `false` arm emits one predicate per keyword
+  * and joins the words *inside* a keyword with ` & ` (rendering `to_tsquery('english', 'alpha & beta')`).
+  * `to_tsquery('english', 'alpha & beta')`). So assert on individual tokens — never on a joined
+  * multi-token string, and never on one arm's own output; either would tie this spec's result to
+  * whichever other suites wrote the flag first.
   *
   * The `record.into(USER).into(classOf[User]).getEmail` in `VersionedResourceTables.hydrate` used to
   * be executed but unobservable: the dataset schema left `UnifiedResourceSchema`'s `userEmail` at its
diff --git a/amber/src/test/scala/org/apache/texera/web/resource/dashboard/file/DatasetResourceSpec.scala b/amber/src/test/scala/org/apache/texera/web/resource/dashboard/file/DatasetResourceSpec.scala
index 70b8252c096..7e58489556b 100644
--- a/amber/src/test/scala/org/apache/texera/web/resource/dashboard/file/DatasetResourceSpec.scala
+++ b/amber/src/test/scala/org/apache/texera/web/resource/dashboard/file/DatasetResourceSpec.scala
@@ -25,7 +25,6 @@ import org.apache.texera.dao.jooq.generated.enums.UserRoleEnum
 import org.apache.texera.dao.jooq.generated.tables.pojos.User
 import org.apache.texera.web.resource.dashboard.DashboardResource.SearchQueryParams
 import org.apache.texera.web.resource.dashboard.user.dataset.DatasetResource.DashboardDataset
-import org.apache.texera.web.resource.dashboard.{FulltextSearchQueryUtils}
 import org.apache.texera.web.resource.dashboard.DatasetSearchQueryBuilder
 import org.scalatest.flatspec.AnyFlatSpec
 import org.apache.texera.dao.jooq.generated.tables.daos.{UserDao, DatasetDao, DatasetUserAccessDao}
@@ -33,7 +32,6 @@ import org.apache.texera.dao.jooq.generated.enums.PrivilegeEnum
 import org.apache.texera.dao.jooq.generated.tables.pojos.{Dataset, DatasetUserAccess}
 import org.scalatest.{BeforeAndAfterAll, BeforeAndAfterEach}
 import java.time.OffsetDateTime
-import java.util
 import org.apache.texera.web.resource.dashboard.SearchQueryBuilder.DATASET_RESOURCE_TYPE
 
 class DatasetResourceSpec
@@ -96,7 +94,14 @@ class DatasetResourceSpec
 
   override protected def beforeAll(): Unit = {
     initializeDBAndReplaceDSLContext()
-    FulltextSearchQueryUtils.usePgroonga = false // disable pgroonga
+    // `FulltextSearchQueryUtils.usePgroonga` is deliberately left at its production default here:
+    // no test in this suite reaches the read, because both `constructQuery` calls below pass no
+    // keywords and `getFullTextSearchFilter` returns before that read on an empty keyword list.
+    // A keyword test added here would need the `to_tsvector` arm instead — `MockTexeraDB` strips
+    // the full-text index block out of the DDL, so the embedded Postgres has no pgroonga extension
+    // — and would have to set the flag and put it back: amber has no `Test / fork`, so a value
+    // left behind here follows every suite scheduled after this one in the same JVM.
+
     // add test user directly
     val userDao = new UserDao(getDSLContext.configuration())
     userDao.insert(ownerUser)
@@ -124,14 +129,6 @@ class DatasetResourceSpec
     shutdownDB()
   }
 
-  private def getKeywordsArray(keywords: String*): util.ArrayList[String] = {
-    val keywordsList = new util.ArrayList[String]()
-    for (keyword <- keywords) {
-      keywordsList.add(keyword)
-    }
-    keywordsList
-  }
-
   private def assertSameDataset(a: Dataset, b: DashboardDataset): Unit = {
     assert(a.getName == b.dataset.getName)
   }