diff --git a/flox/core.py b/flox/core.py
index b32e3b043..c9f15059a 100644
--- a/flox/core.py
+++ b/flox/core.py
@@ -214,7 +214,7 @@ def slices_from_chunks(chunks):
     return product(*slices)
 
 
-# @memoize
+@memoize
 def find_group_cohorts(labels, chunks, merge: bool = True) -> dict:
     """
     Finds groups labels that occur together aka "cohorts"
@@ -312,13 +312,10 @@ def invert(x) -> tuple[np.ndarray, ...]:
 
         # precompute needed metrics for the quadratic loop below.
         items = tuple((k, len(k), set(k), v) for k, v in sorted_chunks_cohorts.items() if k)
-        import copy
-        items2 = list(copy.deepcopy(items[1:]))
 
         merged_cohorts = {}
         merged_keys: set[tuple] = set()
 
-        # import ipdb; ipdb.set_trace()
         # Now we iterate starting with the longest number of chunks,
         # and then merge in cohorts that are present in a subset of those chunks
         # I think this is suboptimal and must fail at some point.
@@ -330,20 +327,12 @@ def invert(x) -> tuple[np.ndarray, ...]:
             new_value = v1
             # iterate in reverse since we expect small cohorts
             # to be most likely merged in to larger ones
-            to_delete = []
-            for idx2, (k2, len_k2, set_k2, v2) in enumerate(reversed(items2)):
-                if v1 == v2 or k1 == k2:
-                    continue
-                if (len(set_k2 & new_key) / len_k2) > 0.75:
-                    new_key |= set_k2
-                    new_value += v2
-                    assert items2[len(items2)-idx2 - 1][0] == k2
-                    to_delete.append(len(items2)-idx2 - 1)
-                    merged_keys.update((k2,))
-            # print(to_delete)
-            for delete in reversed(sorted(to_delete)):
-                del items2[delete]
-
+            for k2, len_k2, set_k2, v2 in reversed(items[idx + 1 :]):
+                if k2 not in merged_keys:
+                    if (len(set_k2 & new_key) / len_k2) > 0.75:
+                        new_key |= set_k2
+                        new_value += v2
+                        merged_keys.update((k2,))
             sorted_ = sorted(new_value)
             merged_cohorts[tuple(sorted(new_key))] = sorted_
             if idx == 0 and (len(sorted_) == nlabels) and (sorted_ == ilabels).all():
@@ -352,7 +341,6 @@ def invert(x) -> tuple[np.ndarray, ...]:
         # sort by first label in cohort
         # This will help when sort=True (default)
         # and we have to resort the dask array
-        print(merged_cohorts)
         return dict(sorted(merged_cohorts.items(), key=lambda kv: kv[1][0]))
 
     else: