diff --git a/acceptance/acceptance_test.go b/acceptance/acceptance_test.go
index 2d3326b23d6..fcfeda7ed20 100644
--- a/acceptance/acceptance_test.go
+++ b/acceptance/acceptance_test.go
@@ -609,6 +609,29 @@ func getSkipReason(config *internal.TestConfig, configPath string) string {
 	return ""
 }
 
+var ciRunID = regexp.MustCompile(`^[0-9]{1,16}$`)
+
+// ciUniqueName embeds a CI run id into the random unique name as "ci<runID>x<random>".
+// The result stays purely lowercase-alphanumeric like the base32 name it replaces, so it
+// remains valid everywhere $UNIQUE_NAME is used: app names (no hyphens would be fine but
+// underscores/uppercase are not), Python and Unity Catalog identifiers (no hyphens). No
+// punctuation separator works for all of them, so the run id (all digits) is delimited by
+// the letter "x", which also keeps the sweep prefix "ci<runID>x" collision-free between
+// runs whose ids share a prefix. Length is preserved ("app-$UNIQUE_NAME" is exactly the
+// 30-char app name maximum). Returns random unchanged when runID is absent, malformed, or
+// too long to leave at least 8 random characters.
+func ciUniqueName(runID, random string) string {
+	if !ciRunID.MatchString(runID) {
+		return random
+	}
+	prefix := "ci" + runID + "x"
+	randLen := len(random) - len(prefix)
+	if randLen < 8 {
+		return random
+	}
+	return prefix + random[:randLen]
+}
+
 func runTest(t *testing.T,
 	dir string,
 	variant int,
@@ -643,6 +666,8 @@ func runTest(t *testing.T,
 
 	id := uuid.New()
 	uniqueName := strings.ToLower(strings.Trim(base32.StdEncoding.EncodeToString(id[:]), "="))
+	// Embed the CI run id, when present, so leaked resources can be attributed to a run and swept by prefix.
+	uniqueName = ciUniqueName(os.Getenv("GITHUB_RUN_ID"), uniqueName)
 	repls.Set(uniqueName, "[UNIQUE_NAME]")
 
 	var tmpDir string
diff --git a/acceptance/unique_name_test.go b/acceptance/unique_name_test.go
new file mode 100644
index 00000000000..181c648b98b
--- /dev/null
+++ b/acceptance/unique_name_test.go
@@ -0,0 +1,27 @@
+package acceptance_test
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestCIUniqueName(t *testing.T) {
+	// 26 lowercase base32 characters, like the generated unique name.
+	random := "osr5mzrrvzb73juixjoviti24y"
+
+	// Run id embedded, same length as input, lowercase-alphanumeric, sweepable prefix.
+	assert.Equal(t, "ci15799017600xosr5mzrrvzb7", ciUniqueName("15799017600", random))
+	assert.Equal(t, "ci1xosr5mzrrvzb73juixjovit", ciUniqueName("1", random))
+
+	// No or invalid run id: unchanged.
+	assert.Equal(t, random, ciUniqueName("", random))
+	assert.Equal(t, random, ciUniqueName("abc123", random))
+	assert.Equal(t, random, ciUniqueName("123 456", random))
+
+	// 15-digit run id still leaves exactly the 8-char random minimum: prefixed.
+	assert.Equal(t, "ci123456789012345xosr5mzrr", ciUniqueName("123456789012345", random))
+
+	// 16-digit run id is too long to leave enough randomness: unchanged.
+	assert.Equal(t, random, ciUniqueName("1234567890123456", random))
+}
diff --git a/tools/sweep_test_resources.py b/tools/sweep_test_resources.py
new file mode 100755
index 00000000000..d17521871fa
--- /dev/null
+++ b/tools/sweep_test_resources.py
@@ -0,0 +1,85 @@
+#!/usr/bin/env python3
+"""Sweep leaked acceptance-test resources by name prefix.
+
+Lists (and with --delete, deletes) warehouses, pipelines and jobs whose names
+start with the given prefix, e.g. the per-run prefix "ci<GITHUB_RUN_ID>x" that
+the acceptance harness embeds into $UNIQUE_NAME on CI cloud runs.
+
+Authentication is taken from the environment (DATABRICKS_HOST, DATABRICKS_TOKEN
+or any other auth supported by the databricks CLI).
+
+Usage:
+    tools/sweep_test_resources.py ci15799017600x           # dry run: list only
+    tools/sweep_test_resources.py ci15799017600x --delete  # delete matches
+"""
+
+import argparse
+import json
+import subprocess
+import sys
+
+
+def run_json(*args):
+    out = subprocess.check_output(["databricks", *args, "--output", "json"], text=True)
+    return json.loads(out) if out.strip() else []
+
+
+def sweep(kind, items, name_of, id_of, delete_args, prefix, delete):
+    failures = 0
+    for item in items:
+        name = name_of(item) or ""
+        if not name.startswith(prefix):
+            continue
+        res_id = str(id_of(item))
+        print(f"{kind}\t{res_id}\t{name}")
+        if delete:
+            try:
+                subprocess.check_call(["databricks", *delete_args, res_id])
+            except subprocess.CalledProcessError as e:
+                print(f"failed to delete {kind} {res_id}: {e}", file=sys.stderr)
+                failures += 1
+    return failures
+
+
+def main():
+    parser = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
+    parser.add_argument("prefix", help="resource name prefix, e.g. ci<GITHUB_RUN_ID>x")
+    parser.add_argument("--delete", action="store_true", help="delete matches (default: list only)")
+    args = parser.parse_args()
+
+    if not args.prefix:
+        parser.error("prefix must not be empty")
+
+    failures = 0
+    failures += sweep(
+        "warehouse",
+        run_json("warehouses", "list"),
+        lambda w: w.get("name"),
+        lambda w: w.get("id"),
+        ["warehouses", "delete"],
+        args.prefix,
+        args.delete,
+    )
+    failures += sweep(
+        "pipeline",
+        run_json("pipelines", "list-pipelines"),
+        lambda p: p.get("name"),
+        lambda p: p.get("pipeline_id"),
+        ["pipelines", "delete"],
+        args.prefix,
+        args.delete,
+    )
+    failures += sweep(
+        "job",
+        run_json("jobs", "list"),
+        lambda j: j.get("settings", {}).get("name"),
+        lambda j: j.get("job_id"),
+        ["jobs", "delete"],
+        args.prefix,
+        args.delete,
+    )
+    return 1 if failures else 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())