diff --git a/swarm/storage/ldbstore.go b/swarm/storage/ldbstore.go
index ae5caf700d..8c18e88a12 100644
--- a/swarm/storage/ldbstore.go
+++ b/swarm/storage/ldbstore.go
@@ -562,9 +562,10 @@ func (s *LDBStore) writeBatches() {
 			log.Error(fmt.Sprintf("spawn batch write (%d entries): %v", b.Len(), err))
 		}
 		close(c)
-		if e >= s.capacity {
+		for e > s.capacity {
 			log.Info("collecting garbage", "entryCnt", e, "capacity", s.capacity)
 			s.collectGarbage(gcArrayFreeRatio)
+			e = s.entryCnt
 		}
 		s.lock.Unlock()
 	}
diff --git a/swarm/storage/memstore_lrucache.go b/swarm/storage/memstore_lrucache.go
index 7e93a5a8e7..a66dcd0c22 100644
--- a/swarm/storage/memstore_lrucache.go
+++ b/swarm/storage/memstore_lrucache.go
@@ -31,6 +31,12 @@ type MemStore struct {
 	disabled bool
 }
 
+//NewMemStore is instantiating a MemStore cache. We are keeping a record of all outgoing requests for chunks, that
+//should later be delivered by peer nodes, in the `requests` LRU cache. We are also keeping all frequently requested
+//chunks in the `cache` LRU cache.
+//
+//`requests` LRU cache capacity should ideally never be reached, this is why for the time being it should be initialised
+//with the same value as the LDBStore capacity.
 func NewMemStore(params *StoreParams, _ *LDBStore) (m *MemStore) {
 	if params.CacheCapacity == 0 {
 		return &MemStore{