fix: Allow Retrying status while preventing RollingOut loop

Final fix for ProgressDeadlineExceeded archival issue:
- markAsProgressing: Only blocks RollingOut, allows Retrying/Succeeded
- Progress deadline check: Skips enforcement for Archived revisions
- This allows archival teardown to show Retrying state without triggering deadline reset
- Prevents original RollingOut reconcile loop for Active revisions

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

Author: camilamacedo86

internal/operator-controller/controllers/clusterobjectset_controller.go

@@ -672,27 +672,27 @@ func setRetryingConditions(cos *ocv1.ClusterObjectSet, message string) {
 // markAsProgressing sets the Progressing condition to True with the given reason.
 //
 // Once ProgressDeadlineExceeded has been set for the current generation, this function
-// keeps that condition sticky for non-terminal progressing reasons that would otherwise
-// cause the reconciler to oscillate between Progressing=True and
-// ProgressDeadlineExceeded=False for the same generation.
+// blocks only the RollingOut reason to prevent a reconcile loop for Active revisions.
+// The loop would occur when reconcile() sees in-transition objects and sets
+// Progressing=True/RollingOut, then the progress deadline check immediately resets it
+// to ProgressDeadlineExceeded, causing a status-only update that triggers another reconcile.
 //
-// In particular, both RollingOut and Retrying are blocked for Active revisions.
-// Allowing those reasons to overwrite ProgressDeadlineExceeded can trigger unnecessary
-// status-only reconciles and interfere with retry/requeue behavior due to the progress
-// deadline check resetting the condition.
+// Archived revisions are exempt because they skip progress deadline enforcement entirely,
+// so their status updates won't be overridden.
 //
-// Archived revisions are exempt from this guard because the progress deadline check
-// does not apply to them, allowing archival teardown retries to show Retrying status.
+// Other reasons like Succeeded and Retrying are always allowed because:
+// - Succeeded represents terminal success (rollout eventually completed)
+// - Retrying represents transient errors that need to be visible in status
 //
-// If the generation changed (spec update) since ProgressDeadlineExceeded was recorded,
-// the guard allows the update through so the condition reflects the new generation.
+// If the generation changed since ProgressDeadlineExceeded was recorded, the guard
+// allows all updates through so the condition reflects the new generation.
 func markAsProgressing(cos *ocv1.ClusterObjectSet, reason, message string) {
 	if cnd := meta.FindStatusCondition(cos.Status.Conditions, ocv1.ClusterObjectSetTypeProgressing); cnd != nil &&
 		cnd.Status == metav1.ConditionFalse && cnd.Reason == ocv1.ReasonProgressDeadlineExceeded &&
 		cnd.ObservedGeneration == cos.Generation &&
-		(reason == ocv1.ReasonRollingOut || reason == ocv1.ClusterObjectSetReasonRetrying) &&
+		reason == ocv1.ReasonRollingOut &&
 		cos.Spec.LifecycleState != ocv1.ClusterObjectSetLifecycleStateArchived {
-		log.Log.V(1).Info("skipping markAsProgressing: ProgressDeadlineExceeded is sticky for non-terminal progressing reasons",
+		log.Log.V(1).Info("skipping markAsProgressing: ProgressDeadlineExceeded is sticky for RollingOut on Active revisions",
 			"requestedReason", reason, "revision", cos.Name, "lifecycleState", cos.Spec.LifecycleState)
 		return
 	}

internal/operator-controller/controllers/clusterobjectset_controller_internal_test.go

@@ -300,7 +300,7 @@ func Test_markAsProgressing_doesNotOverwriteProgressDeadlineExceeded(t *testing.
 			expectReason:      ocv1.ReasonSucceeded,
 		},
 		{
-			name:       "does not overwrite ProgressDeadlineExceeded with Retrying (same generation)",
+			name:       "allows Retrying to overwrite ProgressDeadlineExceeded (same generation)",
 			generation: 1,
 			existingConditions: []metav1.Condition{
 				{
@@ -311,8 +311,8 @@ func Test_markAsProgressing_doesNotOverwriteProgressDeadlineExceeded(t *testing.
 				},
 			},
 			reason:            ocv1.ClusterObjectSetReasonRetrying,
-			expectProgressing: metav1.ConditionFalse,
-			expectReason:      ocv1.ReasonProgressDeadlineExceeded,
+			expectProgressing: metav1.ConditionTrue,
+			expectReason:      ocv1.ClusterObjectSetReasonRetrying,
 		},
 		{
 			name:       "allows RollingOut when generation changed since ProgressDeadlineExceeded",