databricks
diff --git a/.github/workflows/code-coverage.yml b/.github/workflows/code-coverage.yml
@@ -11,6 +11,19 @@ on:
   pull_request:
   workflow_dispatch:
 
+# Serialise E2E runs per ref so a force-push (or a fast follow-up commit)
+# on a PR cancels the previous run instead of racing it against shared
+# warehouse state (Delta tables, UC Volume files, etc.).
+#
+# Pushes to main are NOT cancelled — each merge commit needs its own clean
+# CI signal so a regression on commit N doesn't get hidden by commit N+1
+# arriving seconds later. (Concurrent main runs can still collide on shared
+# state, but that's the cost of preserving per-commit signal; the
+# uuid-suffix conventions in the e2e tests are what keep them isolated.)
+concurrency:
+  group: e2e-${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.event_name == 'pull_request' }}
+
 jobs:
   test-with-coverage:
     runs-on:

diff --git a/tests/e2e/common/uc_volume_tests.py b/tests/e2e/common/uc_volume_tests.py
@@ -1,5 +1,6 @@
 import os
 import tempfile
+from uuid import uuid4
 
 import pytest
 import databricks.sql as sql
@@ -40,12 +41,16 @@ def test_uc_volume_life_cycle(self, catalog, schema):
         with open(fh, "wb") as fp:
             fp.write(original_text)
 
+        # Unique per-run path so concurrent CI jobs sharing the same volume
+        # don't step on each other's PUT/GET/REMOVE.
+        volume_path = f"/Volumes/{catalog}/{schema}/e2etests/life_cycle_{uuid4().hex[:8]}.csv"
+
         with self.connection(
             extra_params={"staging_allowed_local_path": temp_path}
         ) as conn:
 
             cursor = conn.cursor()
-            query = f"PUT '{temp_path}' INTO '/Volumes/{catalog}/{schema}/e2etests/file1.csv' OVERWRITE"
+            query = f"PUT '{temp_path}' INTO '{volume_path}' OVERWRITE"
             cursor.execute(query)
 
         # GET should succeed
@@ -56,7 +61,7 @@ def test_uc_volume_life_cycle(self, catalog, schema):
             extra_params={"staging_allowed_local_path": new_temp_path}
         ) as conn:
             cursor = conn.cursor()
-            query = f"GET '/Volumes/{catalog}/{schema}/e2etests/file1.csv' TO '{new_temp_path}'"
+            query = f"GET '{volume_path}' TO '{new_temp_path}'"
             cursor.execute(query)
 
         with open(new_fh, "rb") as fp:
@@ -66,7 +71,7 @@ def test_uc_volume_life_cycle(self, catalog, schema):
 
         # REMOVE should succeed
 
-        remove_query = f"REMOVE '/Volumes/{catalog}/{schema}/e2etests/file1.csv'"
+        remove_query = f"REMOVE '{volume_path}'"
 
         # Use minimal retry settings to fail fast
         extra_params = {
@@ -84,7 +89,7 @@ def test_uc_volume_life_cycle(self, catalog, schema):
                 Error, match="Staging operation over HTTP was unsuccessful: 404"
             ):
                 cursor = conn.cursor()
-                query = f"GET '/Volumes/{catalog}/{schema}/e2etests/file1.csv' TO '{new_temp_path}'"
+                query = f"GET '{volume_path}' TO '{new_temp_path}'"
                 cursor.execute(query)
 
         os.remove(temp_path)
@@ -151,19 +156,22 @@ def test_uc_volume_put_fails_if_file_exists_and_overwrite_not_set(
         with open(fh, "wb") as fp:
             fp.write(original_text)
 
+        # Unique per-run path so a concurrent CI job's REMOVE doesn't delete
+        # our file between the two PUTs and silently turn the expected
+        # FILE_IN_STAGING_PATH_ALREADY_EXISTS into a successful PUT.
+        volume_path = f"/Volumes/{catalog}/{schema}/e2etests/put_conflict_{uuid4().hex[:8]}.csv"
+
         def perform_put():
             with self.connection(
                 extra_params={"staging_allowed_local_path": temp_path}
             ) as conn:
                 cursor = conn.cursor()
-                query = f"PUT '{temp_path}' INTO '/Volumes/{catalog}/{schema}/e2etests/file1.csv'"
+                query = f"PUT '{temp_path}' INTO '{volume_path}'"
                 cursor.execute(query)
 
         def perform_remove():
             try:
-                remove_query = (
-                    f"REMOVE '/Volumes/{catalog}/{schema}/e2etests/file1.csv'"
-                )
+                remove_query = f"REMOVE '{volume_path}'"
 
                 with self.connection(
                     extra_params={"staging_allowed_local_path": "/"}

diff --git a/tests/e2e/test_transactions.py b/tests/e2e/test_transactions.py
@@ -35,10 +35,17 @@
 
 
 def _unique_table_name(request):
-    """Derive a unique Delta table name from the test node id."""
+    """Derive a unique Delta table name from the test node id.
+
+    The uuid suffix keeps tables unique across concurrent CI jobs that
+    share the same warehouse/catalog — without it, two runs racing on
+    the same test name collide on CREATE/DROP.
+    """
     node_id = request.node.name
     sanitized = re.sub(r"[^a-z0-9_]", "_", node_id.lower())
-    return f"mst_pysql_{sanitized}"[:80]
+    suffix = uuid.uuid4().hex[:8]
+    # Reserve room for the 9-char "_{suffix}" tail so total stays <= 80.
+    return f"mst_pysql_{sanitized}"[:71] + f"_{suffix}"
 
 
 def _unique_table_name_raw(suffix):