diff --git a/acceptance/acceptance_test.go b/acceptance/acceptance_test.go
index e5781f88f43..09f562ae013 100644
--- a/acceptance/acceptance_test.go
+++ b/acceptance/acceptance_test.go
@@ -534,6 +534,9 @@ func testAccept(t *testing.T, inprocessMode bool, singleTest string) int {
 			var extraVars []string
 			if cloudEnv != "" {
 				extraVars = append(extraVars, "CONFIG_Cloud=true")
+				// CONFIG_CloudName carries the normalized cloud (aws/azure/gcp) so
+				// EnvMatrixExclude rules can drop a variant on one cloud only.
+				extraVars = append(extraVars, "CONFIG_CloudName="+getCloudEnvBase(cloudEnv))
 			}
 
 			expanded := internal.ExpandEnvMatrix(config.EnvMatrix, config.EnvMatrixExclude, extraVars)
diff --git a/acceptance/bundle/invariant/configs/postgres_branch.yml.tmpl b/acceptance/bundle/invariant/configs/postgres_branch.yml.tmpl
index e17859b0b90..03aac05eb94 100644
--- a/acceptance/bundle/invariant/configs/postgres_branch.yml.tmpl
+++ b/acceptance/bundle/invariant/configs/postgres_branch.yml.tmpl
@@ -7,8 +7,11 @@ resources:
       project_id: test-pg-project-$UNIQUE_NAME
       display_name: Test Postgres Project
 
+  # Take over the branch implicitly provisioned with the project rather than
+  # creating a new one.
   postgres_branches:
     foo:
       parent: ${resources.postgres_projects.project.name}
-      branch_id: test-branch-$UNIQUE_NAME
+      branch_id: production
+      replace_existing: true
       no_expiry: true
diff --git a/acceptance/bundle/invariant/configs/postgres_endpoint.yml.tmpl b/acceptance/bundle/invariant/configs/postgres_endpoint.yml.tmpl
index fbffd428ff6..3a4a8b9f67f 100644
--- a/acceptance/bundle/invariant/configs/postgres_endpoint.yml.tmpl
+++ b/acceptance/bundle/invariant/configs/postgres_endpoint.yml.tmpl
@@ -7,14 +7,19 @@ resources:
       project_id: test-pg-project-$UNIQUE_NAME
       display_name: Test Postgres Project
 
+  # Creating the project implicitly provisions a "production" branch and its
+  # primary read-write endpoint, so take both over with replace_existing rather
+  # than creating new ones (a second read_write endpoint is rejected with 400).
   postgres_branches:
     branch:
       parent: ${resources.postgres_projects.project.name}
-      branch_id: test-branch-$UNIQUE_NAME
+      branch_id: production
+      replace_existing: true
       no_expiry: true
 
   postgres_endpoints:
     foo:
       parent: ${resources.postgres_branches.branch.name}
-      endpoint_id: test-endpoint-$UNIQUE_NAME
+      endpoint_id: primary
+      replace_existing: true
       endpoint_type: ENDPOINT_TYPE_READ_WRITE
diff --git a/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl b/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl
index 7a548a02756..08c9a265769 100644
--- a/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl
+++ b/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl
@@ -2,15 +2,42 @@ bundle:
   name: test-bundle-$UNIQUE_NAME
 
 resources:
+  # Storage for the synced table's backing pipeline.
+  schemas:
+    pipeline_storage:
+      name: pipeline_storage_$UNIQUE_NAME
+      catalog_name: main
+
+  postgres_projects:
+    project:
+      project_id: test-pg-project-$UNIQUE_NAME
+      display_name: Test Postgres Project
+
+  # Take over the branch implicitly provisioned with the project rather than
+  # creating a new one.
+  postgres_branches:
+    branch:
+      parent: ${resources.postgres_projects.project.name}
+      branch_id: production
+      replace_existing: true
+      no_expiry: true
+
+  postgres_catalogs:
+    catalog:
+      catalog_id: test_pg_catalog_$UNIQUE_NAME
+      branch: ${resources.postgres_branches.branch.name}
+      postgres_database: appdb
+      create_database_if_missing: true
+
   postgres_synced_tables:
     foo:
-      synced_table_id: lakebase_$UNIQUE_NAME.public.trips_synced
-      source_table_full_name: main.raw.trips
-      primary_key_columns: [id]
+      synced_table_id: ${resources.postgres_catalogs.catalog.catalog_id}.public.trips_synced
+      source_table_full_name: main.test_synced_$UNIQUE_NAME.trips_source
+      primary_key_columns: [tpep_pickup_datetime]
       scheduling_policy: SNAPSHOT
       postgres_database: appdb
-      branch: projects/test-pg-project-$UNIQUE_NAME/branches/production
+      branch: ${resources.postgres_branches.branch.name}
       create_database_objects_if_missing: true
       new_pipeline_spec:
-        storage_catalog: main
-        storage_schema: pipelines
+        storage_catalog: ${resources.schemas.pipeline_storage.catalog_name}
+        storage_schema: ${resources.schemas.pipeline_storage.name}
diff --git a/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl-cleanup.sh b/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl-cleanup.sh
new file mode 100755
index 00000000000..4890ef14814
--- /dev/null
+++ b/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl-cleanup.sh
@@ -0,0 +1,6 @@
+#!/bin/bash
+
+# Clean up the temporary source table
+echo "Cleaning up temporary source table"
+$CLI tables delete main.test_synced_$UNIQUE_NAME.trips_source || true
+$CLI schemas delete main.test_synced_$UNIQUE_NAME || true
diff --git a/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl-init.sh b/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl-init.sh
new file mode 100755
index 00000000000..4a47f8aa2bd
--- /dev/null
+++ b/acceptance/bundle/invariant/configs/postgres_synced_table.yml.tmpl-init.sh
@@ -0,0 +1,15 @@
+#!/bin/bash
+
+# Create a unique source table for this test run to avoid hitting the 20-table-per-source limit
+echo "Creating temporary source table: main.test_synced_$UNIQUE_NAME.trips_source"
+
+# Create schema using CLI
+$CLI schemas create test_synced_$UNIQUE_NAME main -o json | jq '{full_name}'
+
+# Create source table from samples.nyctaxi.trips using SQL API
+# MSYS_NO_PATHCONV=1 prevents Git Bash on Windows from converting /api/... to C:/Program Files/Git/api/...
+MSYS_NO_PATHCONV=1 $CLI api post "/api/2.0/sql/statements/" --json "{
+    \"warehouse_id\": \"$TEST_DEFAULT_WAREHOUSE_ID\",
+    \"statement\": \"CREATE TABLE main.test_synced_$UNIQUE_NAME.trips_source AS SELECT * FROM samples.nyctaxi.trips LIMIT 10\",
+    \"wait_timeout\": \"45s\"
+  }" > /dev/null
diff --git a/acceptance/bundle/invariant/continue_293/out.test.toml b/acceptance/bundle/invariant/continue_293/out.test.toml
index 8fd536c2b85..33b84295aa4 100644
--- a/acceptance/bundle/invariant/continue_293/out.test.toml
+++ b/acceptance/bundle/invariant/continue_293/out.test.toml
@@ -27,13 +27,10 @@ EnvMatrix.INPUT_CONFIG = [
   "model_serving_endpoint.yml.tmpl",
   "pipeline.yml.tmpl",
   "pipeline_apply_policy_default_values.yml.tmpl",
-  "postgres_branch.yml.tmpl",
   "postgres_catalog.yml.tmpl",
   "postgres_database.yml.tmpl",
-  "postgres_endpoint.yml.tmpl",
   "postgres_project.yml.tmpl",
   "postgres_role.yml.tmpl",
-  "postgres_synced_table.yml.tmpl",
   "registered_model.yml.tmpl",
   "schema.yml.tmpl",
   "schema_uppercase_name.yml.tmpl",
diff --git a/acceptance/bundle/invariant/continue_293/test.toml b/acceptance/bundle/invariant/continue_293/test.toml
index 902316bea12..6aece425d10 100644
--- a/acceptance/bundle/invariant/continue_293/test.toml
+++ b/acceptance/bundle/invariant/continue_293/test.toml
@@ -26,3 +26,10 @@ EnvMatrixExclude.no_volume_path_job_ref = ["INPUT_CONFIG=volume_path_job_ref.yml
 # The 1000-task scale case is covered by no_drift. Running it here adds ~1.5 min
 # per variant (two full deploys at 1000 tasks) without incremental coverage.
 EnvMatrixExclude.no_pydabs_1000_tasks = ["INPUT_CONFIG=job_pydabs_1000_tasks.yml.tmpl"]
+
+# replace_existing is not supported on v0.293.0: the seed deploy drops the unknown
+# field and then fails trying to create the implicitly-provisioned production branch
+# / primary endpoint. Covered by no_drift and migrate on current CLI.
+EnvMatrixExclude.no_postgres_branch = ["INPUT_CONFIG=postgres_branch.yml.tmpl"]
+EnvMatrixExclude.no_postgres_endpoint = ["INPUT_CONFIG=postgres_endpoint.yml.tmpl"]
+EnvMatrixExclude.no_postgres_synced_table = ["INPUT_CONFIG=postgres_synced_table.yml.tmpl"]
diff --git a/acceptance/bundle/invariant/test.toml b/acceptance/bundle/invariant/test.toml
index bd8f84aa6db..2ea942e4576 100644
--- a/acceptance/bundle/invariant/test.toml
+++ b/acceptance/bundle/invariant/test.toml
@@ -88,20 +88,21 @@ no_alert_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=alert.yml.tmpl"]
 # so this config is local-only (the mock server stores it verbatim).
 no_run_job_ref_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=job_run_job_ref.yml.tmpl"]
 
-# Postgres resources only work on AWS
-no_postgres_project_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=postgres_project.yml.tmpl"]
-no_postgres_branch_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=postgres_branch.yml.tmpl"]
-no_postgres_endpoint_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=postgres_endpoint.yml.tmpl"]
-no_postgres_catalog_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=postgres_catalog.yml.tmpl"]
-no_postgres_synced_table_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=postgres_synced_table.yml.tmpl"]
-no_postgres_role_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=postgres_role.yml.tmpl"]
-no_postgres_database_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=postgres_database.yml.tmpl"]
+# Lakebase v2 (postgres_* resources) is not available on GCP, but works on AWS and Azure,
+# so exclude these on GCP only.
+no_postgres_project_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=postgres_project.yml.tmpl"]
+no_postgres_branch_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=postgres_branch.yml.tmpl"]
+no_postgres_endpoint_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=postgres_endpoint.yml.tmpl"]
+no_postgres_catalog_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=postgres_catalog.yml.tmpl"]
+no_postgres_synced_table_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=postgres_synced_table.yml.tmpl"]
+no_postgres_role_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=postgres_role.yml.tmpl"]
+no_postgres_database_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=postgres_database.yml.tmpl"]
 
-# Lakebase v1 (database instances/catalogs/synced tables) is not available on GCP,
-# so exclude these from all cloud runs (same as the postgres resources above).
-no_database_instance_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=database_instance.yml.tmpl"]
-no_database_catalog_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=database_catalog.yml.tmpl"]
-no_synced_database_table_on_cloud = ["CONFIG_Cloud=true", "INPUT_CONFIG=synced_database_table.yml.tmpl"]
+# Lakebase v1 (database instances/catalogs/synced tables) is not available on GCP, but works on
+# AWS and Azure, so exclude these on GCP only.
+no_database_instance_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=database_instance.yml.tmpl"]
+no_database_catalog_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=database_catalog.yml.tmpl"]
+no_synced_database_table_on_gcp = ["CONFIG_CloudName=gcp", "INPUT_CONFIG=synced_database_table.yml.tmpl"]
 
 # External locations require actual storage credentials with cloud IAM setup
 # which are environment-specific, so we only test locally with the mock server
diff --git a/acceptance/internal/config.go b/acceptance/internal/config.go
index b2d1a9f578f..3d8dd153098 100644
--- a/acceptance/internal/config.go
+++ b/acceptance/internal/config.go
@@ -118,6 +118,7 @@ type TestConfig struct {
 	// For example:
 	//   EnvMatrixExclude.noplantf = ["READPLAN=1", "DATABRICKS_BUNDLE_ENGINE=terraform"]
 	// Special pseudo-var CONFIG_Cloud=true is available for excluding combinations on cloud runs.
+	// Special pseudo-var CONFIG_CloudName=<aws|azure|gcp> is available for excluding combinations on a specific cloud.
 	EnvMatrixExclude map[string][]string
 
 	// List of keys for which to do string replacement value -> [KEY]. If not set, defaults to true.
diff --git a/acceptance/internal/config_test.go b/acceptance/internal/config_test.go
index a81345ed571..87b0d88b960 100644
--- a/acceptance/internal/config_test.go
+++ b/acceptance/internal/config_test.go
@@ -193,6 +193,33 @@ func TestExpandEnvMatrix(t *testing.T) {
 				{"KEY=B"},
 			},
 		},
+		{
+			name: "CONFIG_CloudName excludes variant on matching cloud",
+			matrix: map[string][]string{
+				"INPUT_CONFIG": {"database_instance.yml.tmpl", "job.yml.tmpl"},
+			},
+			exclude: map[string][]string{
+				"no_database_instance_on_gcp": {"CONFIG_CloudName=gcp", "INPUT_CONFIG=database_instance.yml.tmpl"},
+			},
+			extraVars: []string{"CONFIG_Cloud=true", "CONFIG_CloudName=gcp"},
+			expected: [][]string{
+				{"INPUT_CONFIG=job.yml.tmpl"},
+			},
+		},
+		{
+			name: "CONFIG_CloudName keeps variant on non-matching cloud",
+			matrix: map[string][]string{
+				"INPUT_CONFIG": {"database_instance.yml.tmpl", "job.yml.tmpl"},
+			},
+			exclude: map[string][]string{
+				"no_database_instance_on_gcp": {"CONFIG_CloudName=gcp", "INPUT_CONFIG=database_instance.yml.tmpl"},
+			},
+			extraVars: []string{"CONFIG_Cloud=true", "CONFIG_CloudName=aws"},
+			expected: [][]string{
+				{"INPUT_CONFIG=database_instance.yml.tmpl"},
+				{"INPUT_CONFIG=job.yml.tmpl"},
+			},
+		},
 	}
 
 	for _, tt := range tests {