diff --git a/go/test/endtoend/vreplication/fk_test.go b/go/test/endtoend/vreplication/fk_test.go
index a349a94ffa1..5566dcbf6bc 100644
--- a/go/test/endtoend/vreplication/fk_test.go
+++ b/go/test/endtoend/vreplication/fk_test.go
@@ -29,6 +29,7 @@ import (
 	"vitess.io/vitess/go/mysql"
 	"vitess.io/vitess/go/sqltypes"
 	"vitess.io/vitess/go/vt/log"
+	vttablet "vitess.io/vitess/go/vt/vttablet/common"
 
 	binlogdatapb "vitess.io/vitess/go/vt/proto/binlogdata"
 )
@@ -101,6 +102,49 @@ func TestFKWorkflow(t *testing.T) {
 	waitForWorkflowState(t, vc, ksWorkflow, binlogdatapb.VReplicationWorkflowState_Running.String())
 	targetKs := vc.Cells[cellName].Keyspaces[targetKeyspace]
 	targetTab := targetKs.Shards["0"].Tablets[fmt.Sprintf("%s-%d", cellName, targetTabletId)].Vttablet
+
+	// Stop the LoadSimulator while we are testing for workflow error, so that
+	// we don't error out in the LoadSimulator as we will be shutting down source dbServer.
+	if withLoad {
+		cancel()
+		<-ch
+	}
+
+	sourceTab := vc.Cells[cellName].Keyspaces[sourceKeyspace].Shards["0"].Tablets[fmt.Sprintf("%s-%d", cellName, 100)]
+
+	// Stop the source database server to simulate an error during replication phase
+	// This should cause recoverable errors that atomic workflows should retry
+	// as it is already out of copy phase.
+	err := sourceTab.DbServer.Stop()
+	require.NoError(t, err)
+
+	// Give some time for the workflow to encounter errors and potentially retry
+	time.Sleep(2 * vttablet.GetDefaultVReplicationConfig().RetryDelay)
+
+	// Verify workflow is still running and hasn't terminated due to errors
+	waitForWorkflowState(t, vc, ksWorkflow, binlogdatapb.VReplicationWorkflowState_Running.String())
+
+	// Restart the source database to allow workflow to continue
+	err = sourceTab.DbServer.StartProvideInit(false)
+	require.NoError(t, err)
+
+	err = vc.VtctldClient.ExecuteCommand("SetWritable", fmt.Sprintf("%s-%d", cellName, 100), "true")
+	require.NoError(t, err)
+
+	// Restart the LoadSimulator.
+	if withLoad {
+		ctx, cancel = context.WithCancel(context.Background())
+		ls = newFKLoadSimulator(t, ctx)
+		defer func() {
+			select {
+			case <-ctx.Done():
+			default:
+				cancel()
+			}
+		}()
+		go ls.simulateLoad()
+	}
+
 	require.NotNil(t, targetTab)
 	catchup(t, targetTab, workflowName, "MoveTables")
 	vdiff(t, targetKeyspace, workflowName, cellName, nil)
diff --git a/go/vt/vttablet/tabletmanager/vreplication/controller.go b/go/vt/vttablet/tabletmanager/vreplication/controller.go
index 113cb2314a0..e75c4fea0e4 100644
--- a/go/vt/vttablet/tabletmanager/vreplication/controller.go
+++ b/go/vt/vttablet/tabletmanager/vreplication/controller.go
@@ -305,8 +305,8 @@ func (ct *controller) runBlp(ctx context.Context) (err error) {
 		// it's a FAILED_PRECONDITION vterror, OR we cannot identify this as
 		// non-recoverable BUT it has persisted beyond the retry limit
 		// (maxTimeToRetryError). In addition, we cannot restart a workflow
-		// started with AtomicCopy which has _any_ error.
-		if (err != nil && vr.WorkflowSubType == int32(binlogdatapb.VReplicationWorkflowSubType_AtomicCopy)) ||
+		// started with AtomicCopy which has _any_ error during copy phase.
+		if (err != nil && vr.WorkflowSubType == int32(binlogdatapb.VReplicationWorkflowSubType_AtomicCopy) && vr.state == binlogdatapb.VReplicationWorkflowState_Copying) ||
 			isUnrecoverableError(err) ||
 			!ct.lastWorkflowError.ShouldRetry() {
 			err = vterrors.Wrapf(err, TerminalErrorIndicator)