diff --git a/api/containerautomation/autoupdate.go b/api/containerautomation/autoupdate.go
index 782a67ea7..333af7302 100644
--- a/api/containerautomation/autoupdate.go
+++ b/api/containerautomation/autoupdate.go
@@ -177,7 +177,7 @@ func (s *Service) updateEndpoint(endpoint *portainer.Endpoint, scope string, opt
 			continue
 		}
 
-		candidates = append(candidates, UpdateCandidate{ID: c.ID, Name: containerName(c.Names), ImageID: c.ImageID, Labels: c.Labels})
+		candidates = append(candidates, UpdateCandidate{ID: c.ID, Name: containerName(c.Names), ImageID: c.ImageID, Image: c.Image, Labels: c.Labels})
 	}
 
 	// Route and de-duplicate: one redeploy per stack per tick.
@@ -193,7 +193,7 @@ func (s *Service) updateEndpoint(endpoint *portainer.Endpoint, scope string, opt
 	}
 
 	for _, st := range grouped.Stacks {
-		s.updateStack(endpoint, st)
+		s.updateStack(cli, endpoint, st)
 	}
 }
 
@@ -506,7 +506,11 @@ func (s *Service) cleanupOldImage(cli *dockerclient.Client, endpoint *portainer.
 //     change or via a manual "Update now", and we do not fetch git every tick.
 //   - file stacks: the deployer is driven directly with forcePullImage=true,
 //     applying the image update immediately.
-func (s *Service) updateStack(endpoint *portainer.Endpoint, st StackUpdate) {
+//
+// On a successful file-stack redeploy it emits one EventUpdated per member
+// container that triggered the update (not a single aggregate stack event), each
+// carrying the stack name and a best-effort post-redeploy new image id.
+func (s *Service) updateStack(cli *dockerclient.Client, endpoint *portainer.Endpoint, st StackUpdate) {
 	if st.IsGit {
 		// Detect-only: leave git bookkeeping to the git redeploy path. Logged at
 		// debug so it does not repeat at info on every tick (it would otherwise
@@ -553,8 +557,42 @@ func (s *Service) updateStack(endpoint *portainer.Endpoint, st StackUpdate) {
 
 	log.Info().Int("stack_id", st.StackID).Int("endpoint_id", int(endpoint.ID)).
 		Msg("auto-update: redeployed compose stack with updated images")
-	s.notifier.Notify(Event{
-		Kind: EventUpdated, EndpointID: int(endpoint.ID), StackID: st.StackID,
-		Message: "redeployed compose stack with updated images",
-	})
+
+	// One notification PER updated container (the maintainer's requirement), each
+	// showing the container's stack name. The stack was redeployed as a whole, so the
+	// per-container new image id is not in hand; re-inspect each container by its
+	// (compose-stable) name to fill in the "new" digest best-effort. A failed inspect
+	// leaves NewDigest empty and the message falls back to "image updated" — never a
+	// blocked delivery.
+	for _, c := range st.Containers {
+		s.notifier.Notify(Event{
+			Kind: EventUpdated, EndpointID: int(endpoint.ID), StackID: st.StackID,
+			StackName: c.Labels[composeProjectLabel], ContainerName: c.Name,
+			Image: c.Image, OldDigest: c.ImageID, NewDigest: s.inspectImageID(cli, c.Name),
+			Message: "updated stack container",
+		})
+	}
+}
+
+// inspectImageID re-inspects a container by its (compose-stable) name after a stack
+// redeploy to recover the new local image id for the update notification. It is
+// best-effort: any failure (or an empty name) yields "", and the caller degrades the
+// message to "image updated" rather than blocking delivery. The inspect is bounded
+// like every other engine call so a hung engine cannot stall the tick.
+func (s *Service) inspectImageID(cli *dockerclient.Client, containerName string) string {
+	if containerName == "" {
+		return ""
+	}
+
+	ctx, cancel := context.WithTimeout(s.baseCtx, endpointTimeout)
+	defer cancel()
+
+	inspect, err := cli.ContainerInspect(ctx, containerName)
+	if err != nil {
+		log.Debug().Err(err).Str("container", containerName).
+			Msg("auto-update: unable to inspect stack container for its new image id, notifying without it")
+		return ""
+	}
+
+	return inspect.Image
 }
diff --git a/api/containerautomation/autoupdate_test.go b/api/containerautomation/autoupdate_test.go
new file mode 100644
index 000000000..00e80ab1d
--- /dev/null
+++ b/api/containerautomation/autoupdate_test.go
@@ -0,0 +1,148 @@
+package containerautomation
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+
+	portainer "github.com/portainer/portainer/api"
+	"github.com/portainer/portainer/api/datastore"
+	"github.com/portainer/portainer/api/internal/testhelpers"
+
+	"github.com/docker/docker/api/types/container"
+	dockerclient "github.com/docker/docker/client"
+	"github.com/stretchr/testify/require"
+)
+
+// newStackInspectClient builds a Docker client wired to a test server that answers
+// ContainerInspect by name, returning the given new image id. It is the seam the
+// post-redeploy best-effort "new digest" re-inspect uses.
+func newStackInspectClient(t *testing.T, newImageIDByName map[string]string) *dockerclient.Client {
+	t.Helper()
+
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		for name, imageID := range newImageIDByName {
+			if strings.HasSuffix(r.URL.Path, "/containers/"+name+"/json") {
+				_ = json.NewEncoder(w).Encode(container.InspectResponse{
+					ContainerJSONBase: &container.ContainerJSONBase{ID: name, Image: imageID},
+					Config:            &container.Config{},
+				})
+				return
+			}
+		}
+
+		http.Error(w, "not found", http.StatusNotFound)
+	}))
+	t.Cleanup(srv.Close)
+
+	cli, err := dockerclient.NewClientWithOpts(
+		dockerclient.WithHost(srv.URL),
+		dockerclient.WithHTTPClient(http.DefaultClient),
+	)
+	require.NoError(t, err)
+
+	return cli
+}
+
+// TestUpdateStackEmitsPerContainerEvents proves the maintainer's requirement: a
+// (file) stack redeploy emits one EventUpdated PER updated member container, each
+// carrying the compose stack name (from the container's label, not a Stack().Read)
+// and a best-effort post-redeploy new image id — never a single aggregate stack
+// event.
+func TestUpdateStackEmitsPerContainerEvents(t *testing.T) {
+	_, store := datastore.MustNewTestStore(t, true, false)
+
+	// A stack author must exist for registry resolution; an admin resolves to the
+	// (empty) registry set without needing endpoint/team wiring.
+	require.NoError(t, store.User().Create(&portainer.User{ID: 1, Username: "auto", Role: portainer.AdministratorRole}))
+
+	endpoint := &portainer.Endpoint{ID: 1, Name: "nebula.lc"}
+	require.NoError(t, store.Endpoint().Create(endpoint))
+
+	require.NoError(t, store.Stack().Create(&portainer.Stack{
+		ID: 7, EndpointID: 1, Name: "cache-demo", Type: portainer.DockerComposeStack, CreatedBy: "auto",
+	}))
+
+	const (
+		oldEsphome = "sha256:59b94983c73a000000000000000000000000000000000000000000000000aaaa"
+		newEsphome = "sha256:2231ca5d676d000000000000000000000000000000000000000000000000bbbb"
+		oldOther   = "sha256:1111111111110000000000000000000000000000000000000000000000000000"
+		newOther   = "sha256:2222222222220000000000000000000000000000000000000000000000000000"
+	)
+
+	cli := newStackInspectClient(t, map[string]string{
+		"esphome": newEsphome,
+		"other":   newOther,
+	})
+
+	rec := &recordingNotifier{}
+	s := &Service{
+		baseCtx:       context.Background(),
+		dataStore:     store,
+		stackDeployer: testhelpers.NewTestStackDeployer(),
+		notifier:      rec,
+	}
+
+	st := StackUpdate{
+		StackID: 7,
+		IsGit:   false,
+		Containers: []UpdateCandidate{
+			{Name: "esphome", ImageID: oldEsphome, Image: "esphome/esphome:latest", Labels: map[string]string{composeProjectLabel: "cache-demo"}},
+			{Name: "other", ImageID: oldOther, Image: "redis:7", Labels: map[string]string{composeProjectLabel: "cache-demo"}},
+		},
+	}
+
+	s.updateStack(cli, endpoint, st)
+
+	require.Len(t, rec.events, 2, "one EventUpdated per updated member container, not one aggregate stack event")
+
+	byContainer := map[string]Event{}
+	for _, e := range rec.events {
+		require.Equal(t, EventUpdated, e.Kind)
+		require.Equal(t, "cache-demo", e.StackName, "each per-container event carries the compose stack name")
+		require.Equal(t, 7, e.StackID)
+		byContainer[e.ContainerName] = e
+	}
+
+	esphome, ok := byContainer["esphome"]
+	require.True(t, ok, "expected a per-container event for esphome")
+	require.Equal(t, oldEsphome, esphome.OldDigest)
+	require.Equal(t, newEsphome, esphome.NewDigest, "the new image id is recovered by re-inspecting the container after redeploy")
+
+	other, ok := byContainer["other"]
+	require.True(t, ok, "expected a per-container event for other")
+	require.Equal(t, oldOther, other.OldDigest)
+	require.Equal(t, newOther, other.NewDigest)
+}
+
+// TestUpdateStackGitIsDetectOnly guards that a git stack stays detect-only: it is
+// not redeployed and emits no notification (its image update lands on the next git
+// change or a manual update).
+func TestUpdateStackGitIsDetectOnly(t *testing.T) {
+	_, store := datastore.MustNewTestStore(t, true, false)
+
+	endpoint := &portainer.Endpoint{ID: 1, Name: "nebula.lc"}
+	require.NoError(t, store.Endpoint().Create(endpoint))
+
+	deployer := testhelpers.NewTestStackDeployer()
+	rec := &recordingNotifier{}
+	s := &Service{
+		baseCtx:       context.Background(),
+		dataStore:     store,
+		stackDeployer: deployer,
+		notifier:      rec,
+	}
+
+	cli := newStackInspectClient(t, nil)
+
+	s.updateStack(cli, endpoint, StackUpdate{
+		StackID: 9, IsGit: true,
+		Containers: []UpdateCandidate{{Name: "esphome", Labels: map[string]string{composeProjectLabel: "cache-demo"}}},
+	})
+
+	require.Empty(t, rec.events, "a git stack is detect-only, no per-container notification")
+	require.Zero(t, deployer.DeployComposeCallCount, "a git stack must not be redeployed here")
+}
diff --git a/api/containerautomation/labels.go b/api/containerautomation/labels.go
index dff4544bf..d59b48217 100644
--- a/api/containerautomation/labels.go
+++ b/api/containerautomation/labels.go
@@ -161,15 +161,26 @@ type UpdateCandidate struct {
 	ID string
 	// Name is the container's primary name (no leading slash). It is stable across
 	// a recreate and keys the update->rollback loop guard.
-	Name    string
+	Name string
+	// ImageID is the pre-update local image id ("sha256:..."), the "old" digest in a
+	// per-container update notification.
 	ImageID string
-	Labels  map[string]string
+	// Image is the container's image reference (e.g. "nginx:latest"), carried for the
+	// notification.
+	Image  string
+	Labels map[string]string
 }
 
-// StackUpdate identifies a Portainer stack to redeploy once.
+// StackUpdate identifies a Portainer stack to redeploy once, together with the
+// affected member containers so each updated container can emit its own
+// notification (with the stack name) after the redeploy.
 type StackUpdate struct {
 	StackID int
 	IsGit   bool
+	// Containers are the outdated member containers that triggered this stack
+	// redeploy, threaded through from detection so a per-container notification can
+	// be emitted for each (name + old image id + image + labels/stack name).
+	Containers []UpdateCandidate
 }
 
 // GroupedUpdates partitions candidates into their apply paths, de-duplicating
@@ -186,7 +197,10 @@ type GroupedUpdates struct {
 // so a stack with several outdated containers is redeployed only once.
 func groupContainersForUpdate(candidates []UpdateCandidate, stackLookup func(project string) *StackMatch) GroupedUpdates {
 	grouped := GroupedUpdates{}
-	seenStacks := make(map[int]bool)
+	// stackIndex maps a stack id to its slot in grouped.Stacks so a stack is
+	// redeployed once, while every member container is still collected for its own
+	// notification (rather than discarded at the collapse).
+	stackIndex := make(map[int]int)
 
 	for _, c := range candidates {
 		routing := resolveContainerUpdateRouting(c.Labels, stackLookup)
@@ -196,12 +210,14 @@ func groupContainersForUpdate(candidates []UpdateCandidate, stackLookup func(pro
 		case UpdateExternal:
 			grouped.External = append(grouped.External, c)
 		case UpdateStack:
-			if seenStacks[routing.StackID] {
-				continue
+			idx, ok := stackIndex[routing.StackID]
+			if !ok {
+				grouped.Stacks = append(grouped.Stacks, StackUpdate{StackID: routing.StackID, IsGit: routing.IsGit})
+				idx = len(grouped.Stacks) - 1
+				stackIndex[routing.StackID] = idx
 			}
 
-			seenStacks[routing.StackID] = true
-			grouped.Stacks = append(grouped.Stacks, StackUpdate{StackID: routing.StackID, IsGit: routing.IsGit})
+			grouped.Stacks[idx].Containers = append(grouped.Stacks[idx].Containers, c)
 		}
 	}
 
diff --git a/api/containerautomation/labels_test.go b/api/containerautomation/labels_test.go
index cf9759aec..9b130bebc 100644
--- a/api/containerautomation/labels_test.go
+++ b/api/containerautomation/labels_test.go
@@ -195,9 +195,9 @@ func TestGroupContainersForUpdate(t *testing.T) {
 
 	candidates := []UpdateCandidate{
 		{ID: "standalone-1"},
-		{ID: "web-a", Labels: map[string]string{composeProjectLabel: "web"}},
-		{ID: "web-b", Labels: map[string]string{composeProjectLabel: "web"}}, // same stack -> deduped
-		{ID: "api-a", Labels: map[string]string{composeProjectLabel: "api"}},
+		{ID: "web-a", Name: "web-a", Labels: map[string]string{composeProjectLabel: "web"}},
+		{ID: "web-b", Name: "web-b", Labels: map[string]string{composeProjectLabel: "web"}}, // same stack -> deduped redeploy, both kept as members
+		{ID: "api-a", Name: "api-a", Labels: map[string]string{composeProjectLabel: "api"}},
 		{ID: "ext-1", Labels: map[string]string{composeProjectLabel: "unknown"}},
 	}
 
@@ -228,4 +228,21 @@ func TestGroupContainersForUpdate(t *testing.T) {
 	if isGit, ok := got[4]; !ok || !isGit {
 		t.Errorf("stack 4 = (%v, present=%v), want present git stack", isGit, ok)
 	}
+
+	// The stack is redeployed once, but every member container is threaded through
+	// (not discarded at the collapse) so each can emit its own notification.
+	members := map[int][]string{}
+	for _, st := range grouped.Stacks {
+		for _, c := range st.Containers {
+			members[st.StackID] = append(members[st.StackID], c.Name)
+		}
+	}
+
+	if got := members[3]; len(got) != 2 || got[0] != "web-a" || got[1] != "web-b" {
+		t.Errorf("stack 3 members = %v, want [web-a web-b]", got)
+	}
+
+	if got := members[4]; len(got) != 1 || got[0] != "api-a" {
+		t.Errorf("stack 4 members = %v, want [api-a]", got)
+	}
 }
diff --git a/api/containerautomation/notify.go b/api/containerautomation/notify.go
index 231fde7f8..c49b568b8 100644
--- a/api/containerautomation/notify.go
+++ b/api/containerautomation/notify.go
@@ -29,7 +29,13 @@ type Event struct {
 	// by the webhook message. It may be empty for events keyed only by ID.
 	ContainerName string
 	StackID       int
-	Image         string
+	// StackName is the compose project (stack) name a container belongs to, sourced
+	// from its com.docker.compose.project label at detection time. It is set on a
+	// per-container update event for a stack member so the webhook can print a
+	// "Stack [name]" line without a StackID/Stack().Read round-trip; empty for
+	// standalone containers.
+	StackName string
+	Image     string
 	// OldDigest and NewDigest carry the pre/post image identities for an update
 	// (image IDs, e.g. "sha256:59b9..."). They are threaded from the update call
 	// site where they are known and left empty otherwise; the webhook notifier
diff --git a/api/containerautomation/webhook.go b/api/containerautomation/webhook.go
index a8599c38c..9c36cf550 100644
--- a/api/containerautomation/webhook.go
+++ b/api/containerautomation/webhook.go
@@ -144,8 +144,13 @@ func (n webhookNotifier) deliver(webhookURL, message string) {
 func (n webhookNotifier) formatMessage(settings *portainer.Settings, event Event) string {
 	lines := []string{"Environment | " + n.environmentName(event.EndpointID)}
 
-	// Context line: the stack for stack-scoped events, otherwise the container.
+	// Context line: the stack for stack-scoped events, otherwise the container. A
+	// per-container stack-member update carries StackName (from the compose label),
+	// preferred over a StackID/Stack().Read round-trip; the container itself still
+	// names the action line below.
 	switch {
+	case event.StackName != "":
+		lines = append(lines, fmt.Sprintf("Stack [%s]", event.StackName))
 	case event.StackID != 0:
 		lines = append(lines, fmt.Sprintf("Stack [%s]", n.stackName(event.StackID)))
 	case event.ContainerName != "":
diff --git a/api/containerautomation/webhook_test.go b/api/containerautomation/webhook_test.go
index 4610da2dc..ba589863b 100644
--- a/api/containerautomation/webhook_test.go
+++ b/api/containerautomation/webhook_test.go
@@ -205,6 +205,49 @@ func TestFormatMessageStackUpdate(t *testing.T) {
 	}
 }
 
+// TestFormatMessageStackMemberUpdate covers the per-container update of a
+// stack-member container: the context line is the compose stack name (from
+// StackName, no Stack().Read), the action line names the container with its
+// old->new digests. This is the maintainer's target output.
+func TestFormatMessageStackMemberUpdate(t *testing.T) {
+	n, store := newTestWebhookNotifier(t, "unused")
+	createEndpoint(t, store, 1, "nebula.lc")
+
+	settings, _ := store.Settings().Settings()
+
+	msg := n.formatMessage(settings, Event{
+		Kind: EventUpdated, EndpointID: 1, StackID: 7, StackName: "cache-demo",
+		ContainerName: "esphome",
+		OldDigest:     "sha256:59b94983c73aabcd", NewDigest: "sha256:2231ca5d676dabcd",
+	})
+
+	want := "Environment | nebula.lc\nStack [cache-demo]\nUpdate [esphome]: 59b94983c73a → 2231ca5d676d"
+	if msg != want {
+		t.Errorf("got:\n%q\nwant:\n%q", msg, want)
+	}
+}
+
+// TestFormatMessageStackMemberUpdateNoNewDigest covers the best-effort fallback:
+// when the post-redeploy new image id could not be recovered, the message still
+// carries the stack and container and degrades the action line to "image updated"
+// rather than blocking delivery.
+func TestFormatMessageStackMemberUpdateNoNewDigest(t *testing.T) {
+	n, store := newTestWebhookNotifier(t, "unused")
+	createEndpoint(t, store, 1, "nebula.lc")
+
+	settings, _ := store.Settings().Settings()
+
+	msg := n.formatMessage(settings, Event{
+		Kind: EventUpdated, EndpointID: 1, StackID: 7, StackName: "cache-demo",
+		ContainerName: "esphome", OldDigest: "sha256:59b94983c73aabcd",
+	})
+
+	want := "Environment | nebula.lc\nStack [cache-demo]\nUpdate [esphome]: image updated"
+	if msg != want {
+		t.Errorf("got:\n%q\nwant:\n%q", msg, want)
+	}
+}
+
 // TestFormatMessageAutoHeal covers the auto-heal message design.
 func TestFormatMessageAutoHeal(t *testing.T) {
 	n, store := newTestWebhookNotifier(t, "unused")
diff --git a/api/docker/images/status.go b/api/docker/images/status.go
index dabdaf58b..f486b90f1 100644
--- a/api/docker/images/status.go
+++ b/api/docker/images/status.go
@@ -9,6 +9,7 @@ import (
 	"github.com/docker/docker/api/types"
 	"github.com/docker/docker/api/types/container"
 	"github.com/docker/docker/api/types/filters"
+	dockerclient "github.com/docker/docker/client"
 	portainer "github.com/portainer/portainer/api"
 	consts "github.com/portainer/portainer/api/docker/consts"
 
@@ -17,6 +18,7 @@ import (
 	"github.com/pkg/errors"
 	"github.com/rs/zerolog/log"
 	"golang.org/x/sync/errgroup"
+	"golang.org/x/sync/singleflight"
 )
 
 // Status constants
@@ -42,12 +44,32 @@ const (
 	statusCacheTTL            = 5 * time.Minute
 	errorStatusCacheTTL       = 5 * time.Minute
 	maxConcurrentStatusChecks = 8
+
+	// forcedRecheckMinInterval bounds how often a forced re-check (?force=true)
+	// actually contacts the registry for the same local imageID. A manual re-check
+	// bypasses the read caches and issues an outbound registry HEAD; because the
+	// endpoint is available to any env-authorized user with no throttle, an
+	// unbounded loop of forced calls would burn the instance's shared registry
+	// pull-rate quota. Within this window a forced call reuses the just-computed
+	// fresh result instead of re-HEADing. It is tied to the remoteDigestCache TTL so
+	// a genuine manual re-check still gets a fresh answer the first time.
+	forcedRecheckMinInterval = 5 * time.Second
 )
 
 var (
 	statusCache       = cache.New(statusCacheTTL, statusCacheTTL)
 	remoteDigestCache = cache.New(5*time.Second, 5*time.Second)
 	swarmID2NameCache = cache.New(5*time.Second, 5*time.Second)
+
+	// forcedRecheckGroup coalesces concurrent forced re-checks of the SAME local
+	// imageID so N simultaneous ?force=true calls collapse to ONE registry HEAD
+	// (singleflight), rather than each issuing its own outbound HEAD.
+	forcedRecheckGroup singleflight.Group
+	// forcedResultCache holds the last successful forced-recompute result per
+	// imageID for forcedRecheckMinInterval, so rapid successive forced calls reuse it
+	// instead of re-HEADing the registry. Only successful results are stored (errors
+	// are never cached, so a transient failure does not suppress a real re-check).
+	forcedResultCache = cache.New(forcedRecheckMinInterval, forcedRecheckMinInterval)
 )
 
 // Status holds Docker image  analysis
@@ -172,16 +194,30 @@ func (c *DigestClient) containerImageStatus(ctx context.Context, containerID str
 	// and Skipped are cached too (only the error paths return early without caching).
 	//
 	// A forced re-check (force=true) skips this cached read and recomputes against
-	// the registry, then writes the fresh result back into the cache below.
+	// the registry (see forcedImageStatus), then writes the fresh result back into
+	// the cache below.
 	if !force {
 		if s, err := CachedResourceImageStatus(imageID); err == nil {
 			return s, nil
 		}
+
+		return c.computeImageStatus(ctx, cli, imageID, container.Config.Image, false)
 	}
 
+	return c.forcedImageStatus(ctx, cli, imageID, container.Config.Image)
+}
+
+// computeImageStatus performs the full, uncached image-status computation for a
+// resolved local imageID: it parses the configured image reference, inspects the
+// local image for its repo digests/tags, compares them against the remote registry
+// digest (honoring force for the short-lived remoteDigestCache read in checkStatus)
+// and writes the fresh result back into the statusCache. It is the shared body of
+// both the default (cache-miss) and forced-recompute paths; the default path's
+// behaviour is unchanged from the previous inline implementation.
+func (c *DigestClient) computeImageStatus(ctx context.Context, cli *dockerclient.Client, imageID, configImage string, force bool) (Status, error) {
 	digs := make([]digest.Digest, 0)
 	images := make([]*Image, 0)
-	if i, err := ParseImage(ParseImageOptions{Name: container.Config.Image}); err == nil {
+	if i, err := ParseImage(ParseImageOptions{Name: configImage}); err == nil {
 		images = append(images, &i)
 	}
 
@@ -201,13 +237,54 @@ func (c *DigestClient) containerImageStatus(ctx context.Context, containerID str
 
 	s, err := c.checkStatus(ctx, images, digs, force)
 	if err != nil {
-		log.Debug().Str("image", container.Image).Err(err).Msg("fetching a certain image status")
+		log.Debug().Str("image", configImage).Err(err).Msg("fetching a certain image status")
 		return Error, err
 	}
 
 	statusCache.Set(imageID, s, 0)
 
-	return s, err
+	return s, nil
+}
+
+// forcedImageStatus recomputes the image status against the registry for a manual
+// re-check while bounding the registry load a ?force=true call can cause:
+//
+//   - forcedResultCache serves the just-computed fresh result for
+//     forcedRecheckMinInterval, so rapid successive forced calls reuse it instead
+//     of re-HEADing the registry (min-interval throttle).
+//   - forcedRecheckGroup (singleflight) shares one in-flight computation per
+//     imageID, so N concurrent forced calls collapse to a single registry HEAD.
+//
+// A successful recompute still repopulates both the statusCache and (via
+// checkStatus) the remoteDigestCache, so an immediately following default read is
+// served from cache. Failures are not cached, so a transient error never suppresses
+// a genuine re-check.
+func (c *DigestClient) forcedImageStatus(ctx context.Context, cli *dockerclient.Client, imageID, configImage string) (Status, error) {
+	if s, ok := forcedResultCache.Get(imageID); ok {
+		return s.(Status), nil
+	}
+
+	v, err, _ := forcedRecheckGroup.Do(imageID, func() (any, error) {
+		// A concurrent forced call may have populated the result between our read
+		// above and entering the flight; reuse it rather than HEAD the registry again.
+		if s, ok := forcedResultCache.Get(imageID); ok {
+			return s.(Status), nil
+		}
+
+		s, err := c.computeImageStatus(ctx, cli, imageID, configImage, true)
+		if err != nil {
+			return Error, err
+		}
+
+		forcedResultCache.Set(imageID, s, 0)
+
+		return s, nil
+	})
+	if err != nil {
+		return Error, err
+	}
+
+	return v.(Status), nil
 }
 
 func (c *DigestClient) ServiceImageStatus(ctx context.Context, serviceID string, endpoint *portainer.Endpoint) (Status, error) {
diff --git a/api/docker/images/status_test.go b/api/docker/images/status_test.go
index 2aaa54a28..5b404c26c 100644
--- a/api/docker/images/status_test.go
+++ b/api/docker/images/status_test.go
@@ -6,15 +6,19 @@ import (
 	"net/http"
 	"net/http/httptest"
 	"strings"
+	"sync"
+	"sync/atomic"
 	"testing"
 	"time"
 
 	portainer "github.com/portainer/portainer/api"
 
 	"github.com/docker/docker/api/types/container"
+	"github.com/docker/docker/api/types/image"
 	dockerclient "github.com/docker/docker/client"
 	"github.com/opencontainers/go-digest"
 	"github.com/stretchr/testify/require"
+	imagetypes "go.podman.io/image/v5/types"
 )
 
 // fakeClientFactory hands the DigestClient a Docker client wired to a test server.
@@ -131,6 +135,159 @@ func TestCheckStatusForcedBypassesRemoteDigestCache(t *testing.T) {
 	require.Equal(t, Error, forced)
 }
 
+// forcedStubs wires a DigestClient to a fake Docker engine (plain HTTP) and a fake
+// registry (TLS, insecure-skipped) so a forced recompute succeeds end-to-end and
+// resolves an "updated" status. It exposes hit counters for the two expensive calls
+// a forced recompute makes exactly once: the local image inspect (Docker) and the
+// remote manifest HEAD (registry).
+type forcedStubs struct {
+	client       *DigestClient
+	containerID  string
+	imageID      string
+	imageInspect *int64 // Docker ImageInspectWithRaw hits
+	registryHEAD *int64 // registry manifest HEAD hits (the pull-rate-costly call)
+}
+
+// newForcedStubs builds the stubs so the local repo digest matches the remote HEAD
+// digest, i.e. a successful recompute yields images.Updated. registryDelay is slept
+// inside the manifest HEAD handler to widen the window in which concurrent forced
+// calls overlap (used by the singleflight-collapse test).
+func newForcedStubs(t *testing.T, containerID, imageID string, registryDelay time.Duration) forcedStubs {
+	t.Helper()
+
+	const matchDigest = "sha256:" + "cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc"
+
+	var imageInspect, registryHEAD int64
+
+	registry := httptest.NewTLSServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path == "/v2/" {
+			w.WriteHeader(http.StatusOK)
+			return
+		}
+		if strings.Contains(r.URL.Path, "/manifests/") {
+			atomic.AddInt64(&registryHEAD, 1)
+			if registryDelay > 0 {
+				time.Sleep(registryDelay)
+			}
+			w.Header().Set("Docker-Content-Digest", matchDigest)
+			w.Header().Set("Content-Type", "application/vnd.docker.distribution.manifest.v2+json")
+			w.WriteHeader(http.StatusOK)
+			return
+		}
+		http.NotFound(w, r)
+	}))
+	t.Cleanup(registry.Close)
+
+	regHost := strings.TrimPrefix(registry.URL, "https://")
+	imageRef := regHost + "/repo/app:latest"
+
+	docker := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		switch {
+		case strings.HasSuffix(r.URL.Path, "/containers/"+containerID+"/json"):
+			_ = json.NewEncoder(w).Encode(container.InspectResponse{
+				ContainerJSONBase: &container.ContainerJSONBase{ID: containerID, Image: imageID},
+				Config:            &container.Config{Image: imageRef},
+			})
+		case strings.Contains(r.URL.Path, "/images/") && strings.HasSuffix(r.URL.Path, "/json"):
+			atomic.AddInt64(&imageInspect, 1)
+			_ = json.NewEncoder(w).Encode(image.InspectResponse{
+				ID:          imageID,
+				RepoTags:    []string{imageRef},
+				RepoDigests: []string{regHost + "/repo/app@" + matchDigest},
+			})
+		default:
+			http.Error(w, "not found", http.StatusNotFound)
+		}
+	}))
+	t.Cleanup(docker.Close)
+
+	cli, err := dockerclient.NewClientWithOpts(
+		dockerclient.WithHost(docker.URL),
+		dockerclient.WithHTTPClient(http.DefaultClient),
+	)
+	require.NoError(t, err)
+
+	// registryClient nil so RemoteDigest keeps c.sysCtx (insecure skip) rather than
+	// replacing it with an auth-only context; the stub registry uses a self-signed cert.
+	c := &DigestClient{
+		clientFactory: fakeClientFactory{cli: cli},
+		sysCtx:        &imagetypes.SystemContext{DockerInsecureSkipTLSVerify: imagetypes.OptionalBoolTrue},
+	}
+
+	return forcedStubs{
+		client:       c,
+		containerID:  containerID,
+		imageID:      imageID,
+		imageInspect: &imageInspect,
+		registryHEAD: &registryHEAD,
+	}
+}
+
+// TestContainerImageStatusForcedSuccessRepopulatesCache proves the other half of
+// the forced-recheck contract (the error-path tests only cover the bypass): a
+// forced recompute that SUCCEEDS writes the fresh value back into the caches, so a
+// subsequent DEFAULT (non-force) read is served from cache without a second local
+// image inspect or a second registry HEAD.
+func TestContainerImageStatusForcedSuccessRepopulatesCache(t *testing.T) {
+	const containerID = "force-success-container"
+	const imageID = "sha256:0000000000000000000000000000000000000000000000000000000000000002"
+
+	stubs := newForcedStubs(t, containerID, imageID, 0)
+	defer EvictImageStatus(imageID)
+	defer forcedResultCache.Delete(imageID)
+
+	forced, err := stubs.client.ContainerImageStatusForced(context.Background(), containerID, &portainer.Endpoint{}, "")
+	require.NoError(t, err)
+	require.Equal(t, Updated, forced)
+	require.Equal(t, int64(1), atomic.LoadInt64(stubs.imageInspect), "forced recompute inspects the local image once")
+	require.Equal(t, int64(1), atomic.LoadInt64(stubs.registryHEAD), "forced recompute HEADs the registry once")
+
+	// The forced recompute must have repopulated the statusCache under the imageID.
+	cachedStatus, err := CachedResourceImageStatus(imageID)
+	require.NoError(t, err)
+	require.Equal(t, Updated, cachedStatus)
+
+	// A following default read is served from the statusCache: no extra image
+	// inspect and, crucially, no extra registry HEAD.
+	def, err := stubs.client.ContainerImageStatus(context.Background(), containerID, &portainer.Endpoint{}, "")
+	require.NoError(t, err)
+	require.Equal(t, Updated, def)
+	require.Equal(t, int64(1), atomic.LoadInt64(stubs.imageInspect), "default read after a forced success must not re-inspect")
+	require.Equal(t, int64(1), atomic.LoadInt64(stubs.registryHEAD), "default read after a forced success must not re-HEAD the registry")
+}
+
+// TestForcedRechecksCollapseToSingleRegistryHead proves the abuse mitigation: many
+// concurrent forced re-checks of the same imageID collapse (singleflight) to a
+// single outbound registry HEAD instead of one HEAD per call.
+func TestForcedRechecksCollapseToSingleRegistryHead(t *testing.T) {
+	const containerID = "force-collapse-container"
+	const imageID = "sha256:0000000000000000000000000000000000000000000000000000000000000003"
+
+	// A small delay in the manifest HEAD widens the overlap window so the concurrent
+	// callers pile up on the shared in-flight computation.
+	stubs := newForcedStubs(t, containerID, imageID, 100*time.Millisecond)
+	defer EvictImageStatus(imageID)
+	defer forcedResultCache.Delete(imageID)
+
+	const callers = 16
+	var wg sync.WaitGroup
+	wg.Add(callers)
+	for range callers {
+		go func() {
+			defer wg.Done()
+			s, err := stubs.client.ContainerImageStatusForced(context.Background(), containerID, &portainer.Endpoint{}, "")
+			require.NoError(t, err)
+			require.Equal(t, Updated, s)
+		}()
+	}
+	wg.Wait()
+
+	require.Equal(t, int64(1), atomic.LoadInt64(stubs.registryHEAD),
+		"concurrent forced re-checks of the same image must collapse to a single registry HEAD")
+	require.Equal(t, int64(1), atomic.LoadInt64(stubs.imageInspect),
+		"concurrent forced re-checks of the same image must share a single local inspect")
+}
+
 func TestAggregateImageStatus(t *testing.T) {
 	t.Parallel()
 
diff --git a/app/react/docker/containers/ItemView/ContainerActionsSection/SecondaryActions/UpdateNowButton.test.tsx b/app/react/docker/containers/ItemView/ContainerActionsSection/SecondaryActions/UpdateNowButton.test.tsx
index 238e69841..9dc1589b6 100644
--- a/app/react/docker/containers/ItemView/ContainerActionsSection/SecondaryActions/UpdateNowButton.test.tsx
+++ b/app/react/docker/containers/ItemView/ContainerActionsSection/SecondaryActions/UpdateNowButton.test.tsx
@@ -1,4 +1,4 @@
-import { render, screen } from '@testing-library/react';
+import { render, screen, fireEvent, waitFor } from '@testing-library/react';
 import { vi } from 'vitest';
 
 import { COMPOSE_STACK_NAME_LABEL } from '@/react/constants';
@@ -11,11 +11,28 @@ const mockUseImageStatus = vi.fn();
 const mockUseStacks = vi.fn();
 const mockUseAuthorizations = vi.fn();
 const mockMutate = vi.fn();
+const mockConfirmContainerRecreation = vi.fn();
+const mockConfirmStackUpdate = vi.fn();
 
 vi.mock('@uirouter/react', () => ({
   useRouter: () => ({ stateService: { go: vi.fn() } }),
 }));
 
+// The two confirm dialogs are the async gate before the mutation dispatches;
+// stub them so a test can drive the resolved (confirmed) branch.
+vi.mock('@/react/docker/containers/ItemView/ConfirmRecreationModal', () => ({
+  confirmContainerRecreation: (...args: unknown[]) =>
+    mockConfirmContainerRecreation(...args),
+}));
+
+vi.mock('@/react/common/stacks/common/confirm-stack-update', () => ({
+  confirmStackUpdate: (...args: unknown[]) => mockConfirmStackUpdate(...args),
+}));
+
+vi.mock('@/portainer/services/notifications', () => ({
+  notifySuccess: vi.fn(),
+}));
+
 vi.mock('@/react/docker/containers/queries/useContainerImageStatus', () => ({
   useContainerImageStatus: () => mockUseImageStatus(),
 }));
@@ -118,4 +135,60 @@ describe('UpdateNowButton', () => {
     });
     expect(screen.getByTestId('update-now-button')).toBeDisabled();
   });
+
+  // apply() dispatch: clicking confirms, then routes to the mutation exactly once
+  // with the resolved pullImage, via the standalone (recreate) vs stack (redeploy)
+  // confirm dialog.
+  it('applies a standalone update via the recreate confirm, mutating once with pullImage', async () => {
+    setStatus('outdated');
+    mockConfirmContainerRecreation.mockResolvedValue({ pullLatest: true });
+
+    renderButton(); // no compose label -> standalone
+
+    fireEvent.click(screen.getByTestId('update-now-button'));
+
+    await waitFor(() => expect(mockMutate).toHaveBeenCalledTimes(1));
+    expect(mockConfirmContainerRecreation).toHaveBeenCalledTimes(1);
+    expect(mockConfirmStackUpdate).not.toHaveBeenCalled();
+    expect(mockMutate).toHaveBeenCalledWith(
+      expect.objectContaining({ pullImage: true }),
+      expect.anything()
+    );
+  });
+
+  it('applies a stack update via the stack confirm, mutating once with pullImage', async () => {
+    setStatus('outdated');
+    mockUseStacks.mockReturnValue({ data: [composeStack], isLoading: false });
+    mockUseAuthorizations.mockReturnValue({ authorized: true });
+    mockConfirmStackUpdate.mockResolvedValue({ repullImageAndRedeploy: false });
+
+    renderButton({
+      environmentId: 3,
+      labels: { [COMPOSE_STACK_NAME_LABEL]: 'my-stack' },
+    });
+
+    fireEvent.click(screen.getByTestId('update-now-button'));
+
+    await waitFor(() => expect(mockMutate).toHaveBeenCalledTimes(1));
+    expect(mockConfirmStackUpdate).toHaveBeenCalledTimes(1);
+    expect(mockConfirmContainerRecreation).not.toHaveBeenCalled();
+    expect(mockMutate).toHaveBeenCalledWith(
+      expect.objectContaining({ pullImage: false }),
+      expect.anything()
+    );
+  });
+
+  it('does not mutate when the confirm dialog is dismissed', async () => {
+    setStatus('outdated');
+    mockConfirmContainerRecreation.mockResolvedValue(false);
+
+    renderButton();
+
+    fireEvent.click(screen.getByTestId('update-now-button'));
+
+    await waitFor(() =>
+      expect(mockConfirmContainerRecreation).toHaveBeenCalledTimes(1)
+    );
+    expect(mockMutate).not.toHaveBeenCalled();
+  });
 });