diff --git a/pkg/controller/registry/resolver/cache/cache.go b/pkg/controller/registry/resolver/cache/cache.go
index f9b9e03a94..7db59bbc52 100644
--- a/pkg/controller/registry/resolver/cache/cache.go
+++ b/pkg/controller/registry/resolver/cache/cache.go
@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"io"
+	"slices"
 	"sort"
 	"sync"
 	"time"
@@ -147,7 +148,7 @@ func (c *NamespacedOperatorCache) Error() error {
 
 func (c *Cache) Namespaced(namespaces ...string) MultiCatalogOperatorFinder {
 	const (
-		CachePopulateTimeout = time.Minute
+		cachePopulateTimeout = time.Minute
 	)
 
 	sources := c.sp.Sources(namespaces...)
@@ -209,19 +210,38 @@ func (c *Cache) Namespaced(namespaces ...string) MultiCatalogOperatorFinder {
 	}
 	misses = misses[found:]
 
+	// remove any with a "live" outstanding request
+	misses = slices.DeleteFunc(misses, func(key SourceKey) bool {
+		fl := c.logger.(logrus.FieldLogger)
+		fl.Errorf(">>> checking for outstanding request for %s", key)
+
+		hdr := c.snapshots[key]
+
+		// if we already have a request timestamp, we have an outstanding request, so prevent stacking
+		// and just send new requests if the previous one has expired
+		if hdr != nil && hdr.RequestSentinelActive() {
+			fl.Errorf("skipping new request for %s, already in progress", key)
+			return true
+		}
+		return false
+	})
+
 	for _, miss := range misses {
-		ctx, cancel := context.WithTimeout(context.Background(), CachePopulateTimeout)
+		ctx, cancel := context.WithTimeout(context.Background(), cachePopulateTimeout)
 
 		hdr := snapshotHeader{
-			key:      miss,
-			pop:      cancel,
-			priority: c.sourcePriorityProvider.Priority(miss),
+			key:             miss,
+			pop:             cancel,
+			priority:        c.sourcePriorityProvider.Priority(miss),
+			requestSentinel: time.Now().Add(cachePopulateTimeout), // set sentinel to prevent stacking requests
 		}
 
 		hdr.m.Lock()
 		c.snapshots[miss] = &hdr
 		result.snapshots[miss] = &hdr
 
+		// don't adjust the request sentinel in the goroutine for any outcome, so that we don't stampede sources
+		// instead, reevaluate the sentinel during the next snapshot
 		go func(ctx context.Context, hdr *snapshotHeader, source Source) {
 			defer hdr.m.Unlock()
 			c.sem <- struct{}{}
@@ -294,6 +314,8 @@ type snapshotHeader struct {
 	pop      context.CancelFunc
 	err      error
 	priority int
+
+	requestSentinel time.Time
 }
 
 func (hdr *snapshotHeader) Cancel() {
@@ -314,6 +336,13 @@ func (hdr *snapshotHeader) Valid() bool {
 	return true
 }
 
+func (hdr *snapshotHeader) RequestSentinelActive() bool {
+	hdr.m.RLock()
+	defer hdr.m.RUnlock()
+
+	return time.Now().Before(hdr.requestSentinel)
+}
+
 type sortableSnapshots struct {
 	snapshots          []*snapshotHeader
 	preferredNamespace string
diff --git a/test/e2e/util.go b/test/e2e/util.go
index 1241113bec..0db1541315 100644
--- a/test/e2e/util.go
+++ b/test/e2e/util.go
@@ -51,7 +51,7 @@ import (
 
 const (
 	pollInterval = 100 * time.Millisecond
-	pollDuration = 5 * time.Minute
+	pollDuration = 15 * time.Minute
 
 	olmConfigMap = "olm-operators" // No-longer used, how long do we keep this around?