fix: keep load agents connected before start

loadtest/im-pressure/coordinator_test.go

@@ -1,18 +1,57 @@
 package main
 
 import (
+	"bytes"
 	"context"
+	"encoding/json"
 	"net/http"
 	"net/http/httptest"
 	"testing"
 	"time"
 )
 
+func TestCoordinatorResetsBarrierWhenConnectionCountDrops(t *testing.T) {
+	state := &coordinatorState{
+		start: StartInfo{RunID: "run-reset", Scenario: "group"}, startDelay: time.Minute,
+		commitDelay: time.Second, expectedAgents: 2, ready: make(map[string]AgentReady),
+		armed: make(map[string]bool), ledgers: make(map[string][]AcceptedMessage),
+		summaries: make(map[string]AgentSummary), complete: make(chan struct{}),
+	}
+	postReady := func(ready AgentReady) StartInfo {
+		raw, err := json.Marshal(ready)
+		if err != nil {
+			t.Fatal(err)
+		}
+		req := httptest.NewRequest(http.MethodPost, "/v1/ready", bytes.NewReader(raw))
+		w := httptest.NewRecorder()
+		state.handleReady(w, req)
+		if w.Code != http.StatusOK {
+			t.Fatalf("ready status=%d body=%s", w.Code, w.Body.String())
+		}
+		var start StartInfo
+		if err = json.Unmarshal(w.Body.Bytes(), &start); err != nil {
+			t.Fatal(err)
+		}
+		return start
+	}
+	fullA := AgentReady{Machine: "a", ReadyConnections: 1000, ExpectedConnections: 1000}
+	fullB := AgentReady{Machine: "b", ReadyConnections: 1000, ExpectedConnections: 1000}
+	postReady(fullA)
+	if start := postReady(fullB); start.StartAt.IsZero() {
+		t.Fatal("full agents must create a validation barrier")
+	}
+	fullA.ReadyConnections = 999
+	if start := postReady(fullA); !start.StartAt.IsZero() || start.Committed {
+		t.Fatalf("connection drop must reset barrier: %+v", start)
+	}
+}
+
 func TestCoordinatorClientWaitsForBothAgentsAndMergesLedger(t *testing.T) {
 	state := &coordinatorState{
-		start: StartInfo{RunID: "run1", Scenario: "p2p"}, startDelay: time.Second, expectedAgents: 2,
+		start: StartInfo{RunID: "run1", Scenario: "p2p"}, startDelay: 50 * time.Millisecond,
+		commitDelay: 50 * time.Millisecond, expectedAgents: 2,
 		ready: make(map[string]AgentReady), ledgers: make(map[string][]AcceptedMessage),
-		summaries: make(map[string]AgentSummary), complete: make(chan struct{}),
+		armed: make(map[string]bool), summaries: make(map[string]AgentSummary), complete: make(chan struct{}),
 	}
 	server := httptest.NewServer(testCoordinatorMux(state))
 	defer server.Close()
@@ -20,12 +59,16 @@ func TestCoordinatorClientWaitsForBothAgentsAndMergesLedger(t *testing.T) {
 
 	aDone := make(chan StartInfo, 1)
 	go func() {
-		start, _ := client.ready(context.Background(), AgentReady{Machine: "a", ReadyConnections: 1000})
+		start, _ := client.ready(context.Background(), func() AgentReady {
+			return AgentReady{Machine: "a", ReadyConnections: 1000, ExpectedConnections: 1000}
+		})
 		aDone <- start
 	}()
 	time.Sleep(50 * time.Millisecond)
-	startB, err := client.ready(context.Background(), AgentReady{Machine: "b", ReadyConnections: 1000})
-	if err != nil || startB.StartAt.IsZero() {
+	startB, err := client.ready(context.Background(), func() AgentReady {
+		return AgentReady{Machine: "b", ReadyConnections: 1000, ExpectedConnections: 1000}
+	})
+	if err != nil || startB.StartAt.IsZero() || !startB.Committed {
 		t.Fatalf("machine B ready: start=%+v err=%v", startB, err)
 	}
 	select {