fix(e2e): skip stopped node in multinode NATS peer poll

StopNode now stops spinifex-nats.service, so WaitNATSPeers polling that
node's dead 127.0.0.1:8222 monitor port fails with curl exit 7 and never
converges. Add WithSkipNodes to exclude the stopped node from the poll;
TestMultinodeNodeFailure skips node2.

tests/e2e/harness/multinode_cluster.go

@@ -30,6 +30,8 @@ type natsRoutezResponse struct {
 // WaitNATSPeers polls every node's NATS /routez until each reports at least want
 // distinct peers (timeout 60s, interval 2s). NATS monitor binds 127.0.0.1:8222
 // only, so queries run via PeerSSH + curl rather than dialling node.Addr directly.
+// Nodes passed to WithSkipNodes are excluded — use it when a node has been
+// stopped so its dead monitor port doesn't fail the poll.
 func (c *Cluster) WaitNATSPeers(t *testing.T, want int, opts ...PollOpt) {
 	t.Helper()
 	cfg := applyOpts(pollCfg{timeout: 60 * time.Second, interval: 2 * time.Second}, opts...)
@@ -37,6 +39,9 @@ func (c *Cluster) WaitNATSPeers(t *testing.T, want int, opts ...PollOpt) {
 
 	EventuallyErr(t, func() error {
 		for _, n := range c.Nodes {
+			if _, skip := cfg.skipNodes[n.Name]; skip {
+				continue
+			}
 			ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
 			out, err := ssh.Run(ctx, n.Addr,
 				fmt.Sprintf("curl -fsS http://127.0.0.1:%d/routez", natsMonitorPort))

tests/e2e/harness/poll.go

@@ -16,8 +16,9 @@ import (
 type PollOpt func(*pollCfg)
 
 type pollCfg struct {
-	timeout  time.Duration
-	interval time.Duration
+	timeout   time.Duration
+	interval  time.Duration
+	skipNodes map[string]struct{}
 }
 
 // WithTimeout overrides the resource-default timeout.
@@ -26,6 +27,19 @@ func WithTimeout(d time.Duration) PollOpt { return func(c *pollCfg) { c.timeout
 // WithPoll overrides the resource-default polling interval.
 func WithPoll(d time.Duration) PollOpt { return func(c *pollCfg) { c.interval = d } }
 
+// WithSkipNodes excludes named nodes from a per-node cluster poll. Used when a
+// test has intentionally stopped a node (e.g. StopNode) and must not poll it.
+func WithSkipNodes(names ...string) PollOpt {
+	return func(c *pollCfg) {
+		if c.skipNodes == nil {
+			c.skipNodes = make(map[string]struct{}, len(names))
+		}
+		for _, n := range names {
+			c.skipNodes[n] = struct{}{}
+		}
+	}
+}
+
 func applyOpts(def pollCfg, opts ...PollOpt) pollCfg {
 	for _, o := range opts {
 		o(&def)

tests/e2e/multinode/node_failure_test.go

@@ -33,8 +33,9 @@ func runNodeFailure(t *testing.T, fix *Fixture) {
 	})
 
 	// NATS peer-drop detection is delayed by heartbeat; poll so we converge faster than a fixed sleep.
+	// Skip node2 — StopNode stopped its NATS, so its monitor port is dead.
 	harness.Step(t, "wait NATS to report 1 peer (degraded)")
-	fix.Cluster.WaitNATSPeers(t, 1, harness.WithTimeout(30*time.Second), harness.WithPoll(2*time.Second))
+	fix.Cluster.WaitNATSPeers(t, 1, harness.WithTimeout(30*time.Second), harness.WithPoll(2*time.Second), harness.WithSkipNodes(node2.Name))
 
 	harness.Step(t, "DescribeInstanceTypes still answers via %s", local.Name)
 	localCli := harness.AWSClientForGateway(t, fix.Env, local)