diff --git a/cmd/topf/apply.go b/cmd/topf/apply.go
index 636c200..162fd0b 100644
--- a/cmd/topf/apply.go
+++ b/cmd/topf/apply.go
@@ -11,6 +11,7 @@ import (
 	"maps"
 	"slices"
 	"strings"
+	"time"
 
 	"github.com/postfinance/topf/internal/cmd/apply"
 	"github.com/postfinance/topf/internal/nodepool"
@@ -55,6 +56,12 @@ func newApplyCmd() *cli.Command {
 				Sources:     cli.EnvVars("TOPF_ALLOW_NOT_READY"),
 				DefaultText: defaultTextFalse,
 			},
+			&cli.DurationFlag{
+				Name:    "stabilization-duration",
+				Usage:   "how long a node must stay ready after applying before it is considered stable",
+				Value:   time.Second * 30,
+				Sources: cli.EnvVars("TOPF_STABILIZATION_DURATION"),
+			},
 			&cli.StringFlag{
 				Name:    "mode",
 				Value:   "auto",
@@ -83,13 +90,14 @@ func newApplyCmd() *cli.Command {
 			}
 
 			err = apply.Execute(ctx, t, apply.Options{
-				DryRun:               c.Bool("dry-run"),
-				AutoBootstrap:        c.Bool("auto-bootstrap"),
-				SkipProblematicNodes: c.Bool("skip-problematic-nodes"),
-				SkipPostApplyChecks:  c.Bool("skip-post-apply-checks"),
-				AllowNotReady:        c.Bool("allow-not-ready"),
-				Mode:                 mode,
-				MaxParallel:          maxParallel,
+				DryRun:                c.Bool("dry-run"),
+				AutoBootstrap:         c.Bool("auto-bootstrap"),
+				SkipProblematicNodes:  c.Bool("skip-problematic-nodes"),
+				SkipPostApplyChecks:   c.Bool("skip-post-apply-checks"),
+				StabilizationDuration: c.Duration("stabilization-duration"),
+				AllowNotReady:         c.Bool("allow-not-ready"),
+				Mode:                  mode,
+				MaxParallel:           maxParallel,
 			})
 			if errors.Is(err, topf.ErrDryRunChangesDetected) {
 				return cli.Exit(err.Error(), 2)
diff --git a/cmd/topf/upgrade.go b/cmd/topf/upgrade.go
index 5c3fc77..31ad2a0 100644
--- a/cmd/topf/upgrade.go
+++ b/cmd/topf/upgrade.go
@@ -11,6 +11,7 @@ import (
 	"maps"
 	"slices"
 	"strings"
+	"time"
 
 	"github.com/postfinance/topf/internal/cmd/upgrade"
 	"github.com/postfinance/topf/internal/nodepool"
@@ -63,6 +64,12 @@ func newUpgradeCmd() *cli.Command {
 				Sources:     cli.EnvVars("TOPF_DELETE_IF_EVICTION_FAILS"),
 				DefaultText: defaultTextFalse,
 			},
+			&cli.DurationFlag{
+				Name:    "stabilization-duration",
+				Usage:   "how long a node must stay ready after rebooting before it is considered stable",
+				Value:   time.Second * 30,
+				Sources: cli.EnvVars("TOPF_STABILIZATION_DURATION"),
+			},
 			&cli.BoolFlag{
 				Name:        "force",
 				Usage:       "skip etcd health checks during upgrade; only applies to nodes running Talos < 1.13 (legacy MachineService.Upgrade RPC); has no effect on Talos >= 1.13, where the LifecycleService.Upgrade RPC validates etcd health server-side",
@@ -111,6 +118,7 @@ func newUpgradeCmd() *cli.Command {
 				Force:                 c.Bool("force"),
 				Drain:                 c.Bool("drain"),
 				DrainTimeout:          c.Duration("drain-timeout"),
+				StabilizationDuration: c.Duration("stabilization-duration"),
 				DeleteIfEvictionFails: c.Bool("delete-if-eviction-fails"),
 				Stage:                 c.Bool("stage"),
 				StageLabels:           c.StringSlice("stage-label"),
diff --git a/docs/commands/apply.md b/docs/commands/apply.md
index 8520d05..0ccdedc 100644
--- a/docs/commands/apply.md
+++ b/docs/commands/apply.md
@@ -20,10 +20,10 @@ The `apply` command is the primary way to apply configuration changes to a runni
    - Dry-run apply to check for changes
    - If changes detected in `--dry-run` mode: print diff and **exit with code 2**
    - If changes detected in normal mode:
-      - Show diff (if `--confirm` enabled, see [global flags](../configuration.md#global-flags))
-      - Ask for confirmation (if `--confirm` enabled)
-      - Apply configuration
-   - If config applied AND not `--skip-post-apply-checks`: Stabilize (wait 30s for node to be ready)
+     - Show diff (if `--confirm` enabled, see [global flags](../configuration.md#global-flags))
+     - Ask for confirmation (if `--confirm` enabled)
+     - Apply configuration
+   - If config applied AND not `--skip-post-apply-checks`: Stabilize (wait `--stabilization-duration`, default 30s, for node to be ready)
 
 1. **Bootstrap** (if `--auto-bootstrap` enabled):
    - Select first control plane node
@@ -40,7 +40,8 @@ All flags can also be set via environment variables using the `TOPF_` prefix and
 | `--mode`                   | `auto`  | Apply mode: `auto`, `reboot`, `no-reboot`, `staged`, `try`        |
 | `--auto-bootstrap`         | `false` | Automatically bootstrap ETCD after applying configurations         |
 | `--skip-problematic-nodes` | `false` | Continue with healthy nodes if some fail pre-flight checks         |
-| `--skip-post-apply-checks` | `false` | Skip the 30-second stabilization check after applying configs      |
+| `--skip-post-apply-checks` | `false` | Skip the post-apply stabilization check after applying configs   |
+| `--stabilization-duration` | `30s` | How long a node must stay ready after applying before it is considered stable |
 | `--allow-not-ready`        | `false` | Allow applying to nodes that are not ready (have unmet conditions) |
 | [`--nodes-filter`](../configuration.md#filtering-nodes) | - | Regex pattern to filter which nodes to operate on (global flag)    |
 | [`--redact`](../configuration.md#redacting-sensitive-output) | `true` | Redact Talos secrets, certificates, SOPS-encrypted values, and vals-resolved values from output (global flag) |
@@ -73,6 +74,9 @@ topf apply --skip-problematic-nodes
 # Apply without waiting for nodes to stabilize
 topf apply --skip-post-apply-checks
 
+# Apply with a custom stabilization duration
+topf apply --stabilization-duration=1m
+
 # Apply to nodes even if they have unmet conditions
 topf apply --allow-not-ready
 ```
@@ -87,7 +91,7 @@ The apply command validates each node before attempting to apply configuration:
 
 ## Post-apply Stabilization
 
-After applying configuration to a node, the command waits up to 30 seconds for the node to:
+After applying configuration to a node, the command waits for the node to stay ready for `--stabilization-duration` (default: 30 seconds) with:
 
 - Report as ready
 - Have no unmet conditions
diff --git a/docs/commands/upgrade.md b/docs/commands/upgrade.md
index 1af9241..11dd8b9 100644
--- a/docs/commands/upgrade.md
+++ b/docs/commands/upgrade.md
@@ -13,6 +13,7 @@ All flags can also be set via environment variables using the `TOPF_` prefix and
 | `--reboot-mode` | `default` | Reboot mode during upgrade: `default` uses kexec, `powercycle` does a full reboot |
 | `--drain` | `true` | Cordon and drain the Kubernetes node before rebooting, then uncordon after stabilization *(modern flow only; ignored on legacy nodes, where Talos drains and uncordons server-side)* |
 | `--drain-timeout` | `5m` | Maximum time to wait for pod evictions (and, with `--delete-if-eviction-fails`, deletions) to complete during drain *(modern flow only)* |
+| `--stabilization-duration` | `30s` | How long a node must stay ready after rebooting before it is considered stable |
 | `--delete-if-eviction-fails` | `false` | If graceful drain fails (e.g. a PodDisruptionBudget blocks eviction), retry by deleting pods directly (DELETE instead of EVICT, bypassing PDBs); reuses `--drain-timeout` for the delete fallback *(modern flow only)* |
 | `--force` | `false` | Skip etcd health checks; only applies to nodes running Talos < 1.13 (legacy `MachineService.Upgrade` RPC); has no effect on Talos >= 1.13, where the `LifecycleService.Upgrade` RPC validates etcd health server-side |
 | `--stage` | `false` | Install upgrade artifacts without rebooting; the node is left running and can be labeled/annotated/tainted (see `--stage-label`/`--stage-annotation`/`--stage-taint`) so an external controller or human reboots it later |
@@ -78,13 +79,13 @@ All flags can also be set via environment variables using the `TOPF_` prefix and
 1. **If `--stage` is set**: apply labels/annotations/taints (if any) and stop here — the node is not rebooted
 1. Cordon and drain the Kubernetes node if `--drain` is enabled. **If the drain fails** (e.g. a pod cannot be evicted within `--drain-timeout`), the upgrade aborts unless `--delete-if-eviction-fails` is set: in that case, the drain retries with pod deletion (DELETE instead of EVICT, bypassing PodDisruptionBudgets, reusing `--drain-timeout`). If the forced drain also fails, the node is left cordoned with the new artifacts installed but not rebooted, and no further nodes are upgraded. In-flight upgrades on other nodes (when `--max-parallel > 1`) are allowed to complete, but no new ones are started. The node must be uncordoned and rebooted manually to recover.
 1. Issue a `Reboot` with the selected reboot mode (default: kexec)
-1. Wait 30 seconds for the node to stabilize
+1. Wait `--stabilization-duration` (default: 30 seconds) for the node to stabilize
 1. Uncordon the Kubernetes node
 
 **Legacy flow** *(Talos < 1.13)*:
 
 1. Issue `MachineService.Upgrade`, which installs the upgrade artifacts, cordons and drains the node, and reboots — all in a single server-side sequence. `--drain` and `--drain-timeout` are ignored (Talos drains and uncordons the node itself); `--force` skips etcd health checks. `--stage` is not supported on legacy nodes.
-1. Wait 30 seconds for the node to stabilize
+1. Wait `--stabilization-duration` (default: 30 seconds) for the node to stabilize
 
 ## Installer Image
 
diff --git a/internal/cmd/apply/apply.go b/internal/cmd/apply/apply.go
index e5b5087..9ac664d 100644
--- a/internal/cmd/apply/apply.go
+++ b/internal/cmd/apply/apply.go
@@ -30,6 +30,9 @@ type Options struct {
 	SkipPostApplyChecks bool
 	// Allow applying to nodes that are not ready (have unmet conditions)
 	AllowNotReady bool
+	// StabilizationDuration is how long a node must stay ready after
+	// applying before it is considered stable.
+	StabilizationDuration time.Duration
 	// Apply mode passed to Talos (auto, reboot, no-reboot, staged, try)
 	Mode machine.ApplyConfigurationRequest_Mode
 	// MaxParallel controls how many worker nodes are applied to concurrently.
@@ -195,7 +198,7 @@ func applyNode(ctx context.Context, node *topf.Node, opts Options, logger *slog.
 		return nil
 	}
 
-	if err = node.Stabilize(ctx, logger, time.Second*30); err != nil {
+	if err = node.Stabilize(ctx, logger, opts.StabilizationDuration); err != nil {
 		return fmt.Errorf("node didn't stabilize: %w", err)
 	}
 
diff --git a/internal/cmd/upgrade/upgrade.go b/internal/cmd/upgrade/upgrade.go
index 183b838..b9b6c83 100644
--- a/internal/cmd/upgrade/upgrade.go
+++ b/internal/cmd/upgrade/upgrade.go
@@ -57,6 +57,10 @@ type Options struct {
 	// DrainTimeout is the maximum time to wait for pod evictions to complete.
 	DrainTimeout time.Duration
 
+	// StabilizationDuration is how long a node must stay ready after
+	// rebooting before it is considered stable.
+	StabilizationDuration time.Duration
+
 	// DeleteIfEvictionFails retries the drain with direct pod deletion
 	// (DELETE instead of EVICT, bypassing PDBs) if the graceful drain fails.
 	DeleteIfEvictionFails bool
@@ -372,7 +376,7 @@ func upgradeNodeLifecycle(ctx context.Context, t topf.Topf, node *topf.Node, opt
 
 	logger.Info("reboot initiated")
 
-	if err = node.Stabilize(ctx, logger, time.Second*30); err != nil {
+	if err = node.Stabilize(ctx, logger, opts.StabilizationDuration); err != nil {
 		return fmt.Errorf("node didn't stabilize: %w", err)
 	}
 
@@ -420,7 +424,7 @@ func upgradeNodeLegacy(ctx context.Context, _ topf.Topf, node *topf.Node, opts O
 
 	logger.Info("upgrade initiated")
 
-	if err = node.Stabilize(ctx, logger, time.Second*30); err != nil {
+	if err = node.Stabilize(ctx, logger, opts.StabilizationDuration); err != nil {
 		return fmt.Errorf("node didn't stabilize: %w", err)
 	}