diff --git a/tests/test-all.sh b/tests/test-all.sh
index 20eadee2..d7d658c0 100755
--- a/tests/test-all.sh
+++ b/tests/test-all.sh
@@ -38,13 +38,43 @@ run_suite() {
   echo ""
 }
 
-# Start server once
-cd /app/server && bun run src/index.ts &>/dev/null &
-sleep 3
+# Each suite gets a fresh server + fresh DB. Without this, Base E2E leaves
+# 137 tests' worth of state in commhub.db + binds :9200, which then breaks
+# V3 Networks (state pollution → Results line never produced → reported as
+# "0 ran") and Config Priority (its own `bun run src/index.ts &` gets
+# EADDRINUSE because the parent's server still owns :9200, so the suite
+# proceeds against the stale shared server and stops on the first
+# `anet node create` that needs fresh-DB state). See #266 round-1 audit.
+#
+# Suites that don't start their own server (Base E2E, V3 Auth) rely on
+# reset_server() to put one up for them. Suites that DO start their own
+# (V3 Networks, Config Priority) get a free :9200 to bind to.
+reset_server() {
+  pkill -f 'bun.*src/index.ts' 2>/dev/null || true
+  # Wait for :9200 to actually free up — pkill is async; binding before
+  # the old process releases would re-trigger the EADDRINUSE we just fixed.
+  for _ in $(seq 1 30); do
+    if ! (exec 3<>/dev/tcp/127.0.0.1/9200) 2>/dev/null; then break; fi
+    exec 3>&- 2>/dev/null || true
+    sleep 0.25
+  done
+  rm -rf /root/.commhub /root/.anet 2>/dev/null || true
+  cd /app/server && bun run src/index.ts &>/dev/null &
+  for _ in $(seq 1 30); do
+    curl -sf http://127.0.0.1:9200/health > /dev/null && return 0
+    sleep 0.5
+  done
+  echo "::warning::reset_server: server did not respond to /health within 15s" >&2
+  return 1
+}
 
+reset_server
 run_suite "Base E2E (137)" "/app/test.sh 2>&1"
+reset_server
 run_suite "V3 Auth (25)" "/app/test-auth.sh 2>&1"
+reset_server
 run_suite "V3 Networks (22)" "/app/test-networks.sh 2>&1"
+reset_server
 run_suite "Config Priority (16)" "/app/test-config.sh 2>&1"
 
 echo ""