Attempting to fix flaky tests that deal with STUN timeout.

webrtc/p2p/base/stunport_unittest.cc

Index: webrtc/p2p/base/stunport_unittest.cc
diff --git a/webrtc/p2p/base/stunport_unittest.cc b/webrtc/p2p/base/stunport_unittest.cc
index ba4ca364b483bb4ca806579b1cc081f2b7dcb6b4..e5146512df18273f3d39545784511175e0586085 100644
--- a/webrtc/p2p/base/stunport_unittest.cc
+++ b/webrtc/p2p/base/stunport_unittest.cc
@@ -30,7 +30,11 @@ static const SocketAddress kStunAddr3("127.0.0.1", 3000);
 static const SocketAddress kBadAddr("0.0.0.1", 5000);
 static const SocketAddress kStunHostnameAddr("localhost", 5000);
 static const SocketAddress kBadHostnameAddr("not-a-real-hostname", 5000);
-static const int kTimeoutMs = 10000;
+// STUN timeout (with all retries) is 9500ms.
+// Add some margin of error for slow bots.
+// TODO(deadbeef): Use simulated clock instead of just increasing timeouts to
+// fix flaky tests.
+static const int kTimeoutMs = 15000;
 // stun prio = 100 << 24 | 30 (IPV4) << 8 | 256 - 0
 static const uint32_t kStunCandidatePriority = 1677729535;
 static const int kInfiniteLifetime = -1;