diff --git a/pageserver/src/tenant/timeline/compaction.rs b/pageserver/src/tenant/timeline/compaction.rs
index 76dcc159ea..589aea18b4 100644
--- a/pageserver/src/tenant/timeline/compaction.rs
+++ b/pageserver/src/tenant/timeline/compaction.rs
@@ -1112,7 +1112,7 @@ impl Timeline {
         // Accumulate the size of layers in `deltas_to_compact`
         let mut deltas_to_compact_bytes = 0;
 
-        // Under normal circumstances, we will accumulate up to compaction_interval L0s of size
+        // Under normal circumstances, we will accumulate up to compaction_upper_limit L0s of size
         // checkpoint_distance each.  To avoid edge cases using extra system resources, bound our
         // work in this function to only operate on this much delta data at once.
         let delta_size_limit = self.get_compaction_upper_limit() as u64
diff --git a/test_runner/performance/test_compaction.py b/test_runner/performance/test_compaction.py
index 0cd1080fa7..eaa89ae754 100644
--- a/test_runner/performance/test_compaction.py
+++ b/test_runner/performance/test_compaction.py
@@ -75,6 +75,7 @@ def test_compaction_l0_memory(neon_compare: NeonCompare):
             # Initially disable compaction so that we will build up a stack of L0s
             "compaction_period": "0s",
             "gc_period": "0s",
+            "compaction_upper_limit": 12,
         }
     )
     neon_compare.tenant = tenant_id
@@ -91,6 +92,7 @@ def test_compaction_l0_memory(neon_compare: NeonCompare):
     tenant_conf = pageserver_http.tenant_config(tenant_id)
     assert tenant_conf.effective_config["checkpoint_distance"] == 256 * 1024 * 1024
     assert tenant_conf.effective_config["compaction_threshold"] == 10
+    assert tenant_conf.effective_config["compaction_upper_limit"] == 12
 
     # Aim to write about 20 L0s, so that we will hit the limit on how many
     # to compact at once