diff --git a/src/query/interpreter.cpp b/src/query/interpreter.cpp
index 9ba4bc13e..0910ff1f1 100644
--- a/src/query/interpreter.cpp
+++ b/src/query/interpreter.cpp
@@ -2211,7 +2211,10 @@ void RunTriggersIndividually(const utils::SkipList<Trigger> &triggers, Interpret
       const auto &commit_error = maybe_commit_error.GetError();
       switch (commit_error.type) {
         case storage::CommitError::Type::UNABLE_TO_REPLICATE: {
-          spdlog::warn("Unable to replicate to SYNC replica on COMMIT");
+          // TODO(gitbuda): This is tricky because this is an internal
+          // operation. Consider stopping main Memgraph instance here.
+          spdlog::warn("Trigger '{}' failed to commit due to inability to replicate data to SYNC replica",
+                       trigger.Name());
           break;
         }
         case storage::CommitError::Type::CONSTRAINT_VIOLATION: {
@@ -2286,7 +2289,7 @@ void Interpreter::Commit() {
     switch (commit_error.type) {
       case storage::CommitError::Type::UNABLE_TO_REPLICATE: {
         reset_necessary_members();
-        throw QueryException("Unable to replicate to SYNC replica");
+        throw QueryException("Unable to commit due to inability to replicate to SYNC replica");
         break;
       }
       case storage::CommitError::Type::CONSTRAINT_VIOLATION: {
diff --git a/src/storage/v2/storage.cpp b/src/storage/v2/storage.cpp
index 2f67f211c..074c61e13 100644
--- a/src/storage/v2/storage.cpp
+++ b/src/storage/v2/storage.cpp
@@ -1165,6 +1165,10 @@ EdgeTypeId Storage::NameToEdgeType(const std::string_view &name) {
 // because there is no an abort op for that yet, one idea is to just apply
 // reverse operation, e.g., CreateIndex <-> DropIndex.
 //
+// Another idea is to double check that all replicas have relevant data prior
+// to calling MarkFinished. That approach would work in both data replication
+// and global operation cases.
+//
 // EDGE CASE 1: What if the first SYNC replica is alive, receives the delta
 // object, while the second SYNC replica is dead? (replication clients are
 // stored in a vector and accessed one by one)