diff --git a/metropolis/test/util/BUILD.bazel b/metropolis/test/util/BUILD.bazel
index 7caf53b..16ed382 100644
--- a/metropolis/test/util/BUILD.bazel
+++ b/metropolis/test/util/BUILD.bazel
@@ -5,4 +5,5 @@
     srcs = ["runners.go"],
     importpath = "source.monogon.dev/metropolis/test/util",
     visibility = ["//metropolis:__subpackages__"],
+    deps = ["//metropolis/test/launch"],
 )
diff --git a/metropolis/test/util/runners.go b/metropolis/test/util/runners.go
index 47bf59f..a2b1663 100644
--- a/metropolis/test/util/runners.go
+++ b/metropolis/test/util/runners.go
@@ -7,21 +7,25 @@
 	"errors"
 	"testing"
 	"time"
+
+	"source.monogon.dev/metropolis/test/launch"
 )
 
 // TestEventual creates a new subtest looping the given function until it
 // either doesn't return an error anymore or the timeout is exceeded. The last
 // returned non-context-related error is being used as the test error.
 func TestEventual(t *testing.T, name string, ctx context.Context, timeout time.Duration, f func(context.Context) error) {
+	start := time.Now()
 	ctx, cancel := context.WithTimeout(ctx, timeout)
 	t.Helper()
+	launch.Log("Test: %s: starting...", name)
 	t.Run(name, func(t *testing.T) {
 		defer cancel()
 		var lastErr = errors.New("test didn't run to completion at least once")
-		t.Parallel()
 		for {
 			err := f(ctx)
 			if err == nil {
+				launch.Log("Test: %s: okay after %.1f seconds", name, time.Since(start).Seconds())
 				return
 			}
 			if err == ctx.Err() {