diff --git a/core/services/chainlink/application.go b/core/services/chainlink/application.go
index 74168bdd70e..7ac4603b6cc 100644
--- a/core/services/chainlink/application.go
+++ b/core/services/chainlink/application.go
@@ -402,6 +402,13 @@ func NewApplication(ctx context.Context, opts ApplicationOpts) (Application, err
 		srvcs = append(srvcs, durableEmitter)
 	}
 
+	// THROWAWAY: synthetic workflow-event load generator for load-testing the
+	// downstream workflow service from the reliability staging DON. Remove
+	// before this branch is merged. See synthetic_load_emitter.go.
+	if syntheticLoadEnabled {
+		srvcs = append(srvcs, newSyntheticLoadEmitter(globalLogger, syntheticLoadEventsPerSecond))
+	}
+
 	creServices, err := cre.NewServices(
 		globalLogger,
 		opts.DS,
diff --git a/core/services/chainlink/synthetic_load_emitter.go b/core/services/chainlink/synthetic_load_emitter.go
new file mode 100644
index 00000000000..9138e7fb7ee
--- /dev/null
+++ b/core/services/chainlink/synthetic_load_emitter.go
@@ -0,0 +1,78 @@
+package chainlink
+
+// THROWAWAY: synthetic workflow-event load generator.
+//
+// This service emits BaseMessages through the node's global beholder emitter at
+// a fixed, hardcoded rate so we can load-test the downstream workflow service
+// from the reliability staging DON. Because it goes through the standard
+// beholder emitter, every message automatically carries the node's telemetry
+// resource attributes (e.g. platformEnv=staging-reliability) as Kafka headers,
+// which is exactly what the downstream IGNORE_DONS filter matches on.
+//
+// This whole file (and its wiring in application.go) is meant to be deleted
+// before this branch is ever merged.
+
+import (
+	"context"
+	"time"
+
+	commonservices "github.com/smartcontractkit/chainlink-common/pkg/services"
+	"github.com/smartcontractkit/chainlink-common/pkg/timeutil"
+
+	"github.com/smartcontractkit/chainlink-common/pkg/custmsg"
+	"github.com/smartcontractkit/chainlink/v2/core/logger"
+)
+
+// Tuning knobs for the throwaway load generator. Edit and redeploy to change
+// the load; there is intentionally no TOML/env wiring.
+const (
+	// syntheticLoadEnabled turns the generator on. Keep false on any branch
+	// that is not the dedicated reliability load-test deploy.
+	syntheticLoadEnabled = true
+	// syntheticLoadEventsPerSecond is the number of BaseMessages emitted every
+	// second.
+	syntheticLoadEventsPerSecond = 1000
+)
+
+type syntheticLoadEmitter struct {
+	commonservices.Service
+	eng *commonservices.Engine
+
+	emitter         custmsg.MessageEmitter
+	eventsPerSecond int
+}
+
+func newSyntheticLoadEmitter(lggr logger.Logger, eventsPerSecond int) *syntheticLoadEmitter {
+	s := &syntheticLoadEmitter{
+		emitter:         custmsg.NewLabeler().With("source", "synthetic-load"),
+		eventsPerSecond: eventsPerSecond,
+	}
+
+	s.Service, s.eng = commonservices.Config{
+		Name:  "SyntheticLoadEmitter",
+		Start: s.start,
+	}.NewServiceEngine(lggr)
+	return s
+}
+
+func (s *syntheticLoadEmitter) start(_ context.Context) error {
+	if s.eventsPerSecond <= 0 {
+		s.eng.Warnw("synthetic load emitter started with non-positive rate; nothing will be emitted", "eventsPerSecond", s.eventsPerSecond)
+		return nil
+	}
+
+	// Emit the full per-second budget once per tick. A one-second tick keeps the
+	// rate predictable regardless of ticker precision; the burst is fine for a
+	// load test.
+	tick := func(ctx context.Context) {
+		for i := 0; i < s.eventsPerSecond; i++ {
+			if err := s.emitter.Emit(ctx, "synthetic load test event"); err != nil {
+				s.eng.Errorw("synthetic load emit failed", "err", err, "emitted", i)
+				return
+			}
+		}
+	}
+
+	s.eng.GoTick(timeutil.NewTicker(func() time.Duration { return time.Second }), tick)
+	return nil
+}